Group A Streptococcus (M1T1 strain 5448) was utilized for these experiments. Strains included a wild-type M1T1 5448 strain (referred to here as M1 5448), the isogenic mutant sagAΔcat (a gift from V. Nizet, referred to here as the ΔsagA strain), and a complemented sagAΔcat + sagA strain (referred to here as the ΔsagA + sagA strain). The wild-type and ΔsagA strains have been characterized elsewhere (Kansal et al., 2000; Datta et al., 2005). The ΔsagA + sagA strain was generated in our laboratory, and has also been previously described (Flaherty et al., 2015; Higashi et al., 2016). All GAS strains were grown on Todd-Hewitt (TH, Neogen) media or Group A Streptococcus-selective media with 5% sheep blood (BD), and liquid cultures were grown for 16-20 hours at 37° C in TH broth prior to infection experiments. Cultures of the ΔsagA + sagA strain were carried out along with 5 µg/mL erythromycin for selection.

The immortalized human keratinocyte cell line HaCaT (a gift from V. Nizet) was used for these studies, and has been described elsewhere (Boukamp et al., 1988). Cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Gibco) supplemented with 10% heat-inactivated and fetal bovine serum (FBS, Biowest). Prior to supplementation, the FBS was filter-sterilized (0.22 µm Steriflip filters, Millipore). Cells were grown at 37° C with 5% CO₂.

HaCaT cells were seeded onto either sterilized glass coverslips in 6-well tissue culture-treated plates (Corning) for IF microscopy experiments, 35 mm glass bottom imaging dishes (MatTek) for live imaging experiments, or 24-well tissue culture-treated plates (Eppendorf) for cytotoxicity assays and other keratinocyte treatments. Cells were grown to 80-90% confluency, and were washed with sterile 1x PBS (Gibco) prior to the addition of fresh DMEM + 10% FBS. For infections that were performed with chemical inhibitors, this fresh media contained either 100 µM of the compound, or DMSO as a vehicle control. The final DMSO concentration in all wells was 0.8%. Cells were pre-treated for 1 hour at 37° C with 5% CO₂. Meanwhile, overnight GAS cultures were centrifuged and re-suspended in fresh TH, and their optical densities were normalized. Following the pretreatment step, HaCaT cells were infected at a multiplicity of infection (MOI) of 10 for cytotoxicity and IF microscopy experiments, and MOI 0.01 for live imaging experiments. Sterile TH was used as an uninfected control. Infections were carried out at 37° C with 5% CO₂ for the specified times. A summary of the general infection protocol is shown in Figure S1 .

HaCaT cells were infected as described above. Following the infection, supernatants were collected to ensure that any dead, non-adherent cells would be included in the cytotoxicity measurements. Cells remaining on the plate were washed twice with sterile 1x PBS, and washes were pooled with the supernatants. Pooled supernatants and washes were centrifuged at 9,300 x g for 10 minutes to pellet cell debris, and the supernatant was aspirated. This pellet was re-suspended in 4 µM ethidium homodimer 1 (Molecular Probes) in PBS, and was redistributed to a 24-well plate. Cells remaining on the original plate were incubated with the same concentration of ethidium homodimer 1. Both plates were covered and incubated at room temperature with gentle agitation for 30 minutes, and fluorescence was measured on a Synergy H1 Microplate reader set to 528 nm excitation and 617 nm emission, with a cutoff value of 590 nm. To normalize the fluorescence reading to the total number of cells per well, cells were covered again and incubated with 0.1% (wt/vol) Saponin (Sigma) for 20 minutes with gentle agitation, followed by another plate reader measurement using the same settings. The percent membrane permeabilization and corresponding cytotoxicity were calculated using the combined post-infection readings divided by the combined post-Saponin readings for each well.

SLS preparations were generated as previously described (Higashi et al., 2016). Briefly, preparations were generated by transferring overnight GAS cultures on TH plates to TH broth, and optical densities were normalized. 5 x 10⁷ CFU/mL of bacteria were mixed in PBS on ice at a final concentration of 1:40, and were incubated for 30 minutes. All subsequent steps were carried out on ice or at 4° C. Following incubation, bacterial suspensions in PBS were centrifuged at 2000 x g for 10 minutes, and the supernatants were filter sterilized using 0.22 µm Steriflip filters (Millipore). Supernatants were concentrated tenfold by centrifugation using a filtration unit with a 3K MWCO (Pall), and were either used immediately for hemolysis assays or stored at -20°C until use.

Hemolysis assays were performed as previously described (Higashi et al., 2016). Defibrinated whole sheep blood (Lampire) was washed with cold PBS, and was diluted 1:200 in cold PBS. Blood was treated with either DMSO as a vehicle control (final concentration 0.4%) or 50 µM S0859, prior to treatment with SLS preparations at a 1:10 ratio of SLS preparations, PBS, or 10% Triton X-100 (Sigma) in PBS to blood. These mixtures were incubated at 37°C for 1 hour, and were centrifuged at 0.4 x g at 4°C for 10 minutes to pellet intact erythrocytes. Supernatants containing hemoglobin that was released by hemolysis were subjected to absorbance measurements (450 nm), and hemolysis values were calculated by normalizing to the 10% Triton and PBS conditions.

A protein BLAST (NCBI) (Altschul et al., 1997) was used to identify proteins with significant homology to a synthetic peptide that corresponded to a putative binding domain for SLS (query sequence: PWRMHIFTIIQVACLVLLWVVRSIK) (Higashi et al., 2016). This sequence was derived from the C-terminal domain of Band 3 from Ovis aries (NCBI Reference Sequence XP_004013030.2), which was chosen because these Band 3 – SLS studies were performed with sheep blood (Higashi et al., 2016). The search was limited to human proteins (taxid: 9606). Human NBCn1 (solute carrier family 4 sodium bicarbonate cotransporter member 7, GenBank ACH61959.1) was identified as having high similarity to the query, and was aligned with full-length Ovis aries Band 3 using Clustal Omega (Sievers et al., 2011). The Human Protein Atlas (proteinatlas.org) was used to assess expression profiles of selected proteins (Uhlén et al., 2015).

HaCaT cells were seeded onto glass coverslips, treated with inhibitors, and infected with GAS as described above. After the infection, cells were washed with cold PBS and fixed in 4% (wt/vol) paraformaldehyde in PBS for 1 hour. Cells were washed with cold PBS and incubated with a blocking buffer containing 1% (wt/vol) normal goat serum (Invitrogen), 2% (vol/vol) Triton X-100 (Dow/Sigma Aldrich), and 0.5% (vol/vol) Tween 20 (Sigma) in PBS for 2 hours at room temperature. Cells were washed with 1x PBS for 20 minutes and incubated with 1:50 of primary antibody in blocking buffer at 4° C overnight. Coverslips were washed in PBS for 90 minutes, and were incubated with 1:200 secondary antibody in blocking buffer for 2 hours at room temperature. Cells were washed with PBS for 1 hour, and were incubated with 1:500 DAPI nuclear stain and 1:500 rhodamine-phalloidin actin stain in blocking buffer for 30 minutes at room temperature. Coverslips were washed in PBS for 45 minutes, and were mounted to glass slides using Fluoromount-G (Southern Biotech). Once set, slides were imaged using an inverted Nikon Eclipse Ti-E microscope with an iXon Ultra 897 electron multiplying charge-coupled device (Andor) or a Neo sCMOS (Andor) using the 60x objective with oil immersion. Images were analyzed and reconstructed using FIJI/ImageJ (NIH).

Keratinocytes were loaded with pHrodo Red AM (Molecular Probes) according to the manufacturer’s instructions. Briefly, pHrodo Red AM was diluted 10x in PowerLoad concentrate, and this mixture was diluted 100x in Live Cell Imaging Solution (LCIS, Invitrogen). Meanwhile, HaCaT cells were washed 1x with sterile LCIS, and were incubated with the diluted pHrodo dye for 30 minutes while covered at 37°C with 5% CO₂. Excess dye was aspirated, and cells were washed 2x with sterile LCIS. For real-time live imaging experiments, cells were incubated with pre-warmed DMEM + 10% FBS and infected with GAS at an MOI of 0.01. This lower MOI was used to ensure that cells would not be overwhelmed with bacteria over the time course. Infections were imaged on the Nikon Eclipse Ti-E microscope described above, fitted with an environmental chamber set to maintain 37°C and 5% CO₂. Images were taken in the DIC and pHrodo (excitation 555 nm, emission 580 nm, exposure 10 ms) every 10 minutes for 8 hours. Images were reconstructed into time-lapse videos using ImageJ. Experiments were performed at least in triplicate for each condition to confirm individual phenotypes, and representative videos are shown.

For all experiments, human plasminogen-expressing C57BL/6 [hPg (Tg)] (Sun et al., 2004) mice at 6-10 weeks of age were injected subcutaneously in the flank with 100 µL of GAS (10⁷ CFU) with specified concentrations of S0859 or 0.7% saline as a vehicle control. Experiments were performed in a non-blind setting, with no exclusion of animals. Animals for each experimental group were selected randomly. At the end of each experiment, wound measurements and images were taken, and mice were sacrificed under rodent cocktail anesthesia (9 parts ketamine (100 mg/mL) + 9 parts xylazine (20 mg/mL) + 3 parts acepromazine (10 mg/mL) + 79 parts saline), (body weight x 10) - 50 μL = μL/mouse). Wounds were excised and prepared for determination of bacterial colony-forming units as previously described (Datta et al., 2005). Animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Notre Dame.

Two experimental approaches were used: (1) a single dose of treatment (100 µM S0859 or vehicle) administered at the time of bacterial inoculation. Here, 4 mice per group (wild-type + vehicle, wild-type + 100 µM S0859, ΔsagA + vehicle, ΔsagA + sagA + vehicle) were injected in a single experiment. Mice were monitored and wounds were measured every 24 hours for 72 hours. (2) a dose of treatment (200 µM S0859 or vehicle) administered at the time of bacterial inoculation, followed by additional doses at 4 and 8 hours post-infection. Here, the data from two separate experiments were pooled, using only wild-type + vehicle (total n = 7) and wild-type + 200 µM S0859 (total n = 9) as groups. Wound measurements and images were taken after 24 hours, and mice were sacrificed and prepared as described above.

Inhibitor compounds were dissolved in DMSO and used at the indicated concentrations. The N-cyanosulfonamide S0859 (Cayman) and stilbenedisulfonates 4,4’-diisothiocyanatostilbene-2,2’-disulfonate (DIDS, Sigma), 4,4’-diaminostilbene-2,2’-disulfonate (DADS, Sigma), and 4-acetamido-4′-isothiocyanato-stilbene-2,2′-disulfonate (SITS, Sigma) were commercially available, while the DIDS analog 4,4’-ocatanamidostilbene-2,2’-disulfonate (OADS) was synthesized by the Ashfeld group as previously described (Howery et al., 2012). Antibodies to NF-κB (Rb α NF-κB p65, Cell Signaling Technology D14E12, RRID:AB_10859369), rabbit IgG (Gt α Rb IgG AlexaFluor 488, Invitrogen, RRID:AB_143165), and SLC4A7 (Rb α SLC4A7, Fisher PA557433, RRID:AB_2647530) were used at the indicated concentrations. The nuclear stain 4′,6-diamidino-2-phenylindole (DAPI, Cell Signaling Technology) and actin stain rhodamine-phalloidin (Thermo-Fisher) were used at the indicated concentrations.

All experiments were performed at least in triplicate. Statistical analysis was performed using GraphPad Prism 9.0. The similarity of standard deviations between conditions within each experiment and the lack of outlier values led us to assume normal distributions for all data sets. Significance was determined using ANOVA for all experiments in which >2 means were being compared, with post hoc Tukey’s HSD tests as necessary (α = 0.05). Significance was reported as follows for all data sets: *: 0.01 < P < 0.05, **: 0.001 < P < 0.01, ***: 0.0001 < P < 0.001, ****: P < 0.0001, n.s.: not significant.

Previous studies have demonstrated that Band 3-mediated hemolysis induced by SLS can be mitigated by chemically treating erythrocytes with 4,4’-diisothiocyanatostilbene-2,2’-disulfonate (DIDS) prior to the application of bacterial supernatants (Higashi et al., 2016). DIDS is a known Band 3 inhibitor (Okubo et al., 1994; Matulef et al., 2008; Arakawa et al., 2015), and this demonstrated that SLS is capable of targeting host proteins to induce cytolysis. The chemical structure of DIDS is shown in Figure 1A . Although keratinocytes do not express Band 3, we hypothesized that treating keratinocytes with DIDS during a GAS infection would similarly inhibit SLS-dependent cytotoxicity. Therefore, we performed keratinocyte infections following pre-treatment of HaCaT human keratinocytes with 100 µM DIDS, using DMSO as a vehicle control ( Figure 1 ). Following the infection, percent membrane permeabilization corresponding to percent cytotoxicity was assessed using the membrane-impermeable, DNA-intercalating fluorescent dye ethidium homodimer. Keratinocytes were infected for 6 hours at an MOI of 10 based on previous results for optimizing SLS expression in our in vitro infection model (Flaherty et al., 2015).

When HaCaT cells were treated with DMSO as a vehicle control, infection with SLS-producing strains of M1T1 5448 GAS (M1 5448, ΔsagA + sagA) resulted in nearly 50 percent cytotoxicity, which was significantly greater than the cytotoxicity induced by the uninfected control or cells infected with the SLS-deficient ΔsagA strain ( Figure 1B ). These data were consistent with previous studies of GAS infections in the HaCaT cell line (Flaherty et al., 2015). Treatment with 100 µM DIDS did not significantly impact cell viability in the uninfected or ΔsagA conditions, but the cytotoxicity observed in the SLS-producing conditions was reduced to levels comparable to that of the uninfected control. These results demonstrated that the stilbenedisulfonate DIDS inhibits keratinocyte cytotoxicity in an SLS-dependent manner during GAS infection. In addition, other related stilbene derivatives also inhibited SLS-dependent cytotoxicity in keratinocytes during GAS infection ( Figure S2 ). Treatment of keratinocytes with 100 µM each of OADS (4,4’-ocatanamidostilbene-2,2’-disulfonate, Figure S2A ) and SITS (4-acetamido-4′-isothiocyanato-stilbene-2,2′-disulfonate, Figure S2B ), but not DADS (4,4’-diaminostilbene-2,2’-disulfonate, Figure S2C ), inhibited SLS-dependent cytotoxicity in a manner similar to DIDS. DIDS contains two isothiocyanate groups, which are replaced with amines in DADS and alkyl chains connected to the stilbene core through amide linkages in OADS. SITS retains one isothiocyanate from DIDS, while the other is replaced by an acetyl group. This suggested that the isothiocyanate groups of these compounds played a significant role in the inhibition of cytotoxicity, which could be explained by the reported ability of these groups to react with lysine side chains in Band 3 (Okubo et al., 1994).

While the inhibition of SLS-dependent cell death with DIDS and related compounds implicated ion transporters as important players during GAS infections, identifying possible protein targets for SLS in keratinocytes was complicated by the broad inhibitory activity of DIDS and other stilbene derivatives (Matulef and Maduke, 2005; Wulff, 2008; Boedtkjer et al., 2012; Chen et al., 2014). Therefore, we undertook a bioinformatics-based approach to identify plausible keratinocyte targets for SLS. We previously identified a putative binding region for SLS in Band 3 by using a synthetic peptide corresponding to amino acids 848-872 of Band 3 to inhibit SLS-dependent hemolysis (Higashi et al., 2016). Taking into account that this peptide likely inhibited hemolysis by competing with full-length, membrane-bound Band 3 for SLS binding, we used NCBI BLAST to identify other human proteins with similarity to the synthetic peptide sequence. This search largely identified members of the SLC4 family of bicarbonate transporters, which comprise ten genes encoding chloride/bicarbonate exchangers (including Band 3), a borate exchanger, and sodium-coupled bicarbonate transporters (Boron et al., 2009; Romero et al., 2013).

Bioinformatics analysis using this inhibitory peptide identified multiple members of the SLC4 protein family, of which Band 3 is a member (encoded by SLC4A1). Several of the proteins that exhibited the strongest E-value and percent identity were isoforms of the electroneutral sodium-bicarbonate cotransporter 1 (NBCn1, encoded by SLC4A7). These isoforms shared 66.67% identity with the query sequence, with an E-value of 2e-09 (data not shown), indicating that this region is shared across isoforms of NBCn1. Alignment of full-length Band 3 and NBCn1 indicated that the C-terminal region of Band 3 was highly similar to a C-terminal region of NBCn1, and this similarity was especially strong within the region spanned by the synthetic peptide ( Figure 1C ). NBCn1 is an integral membrane protein that is predominantly associated with the maintenance of intracellular pH (pH_i) by moving bicarbonate across the cell membrane for the net effect of either acid loading or acid extrusion (Pushkin et al., 1999; Choi et al., 2000; Boron et al., 2009; Romero et al., 2013; Thornell and Bevensee, 2015). This function is related to that of Band 3, which is a chloride-bicarbonate exchanger that is also involved in ion exchange and maintenance of osmotic balance in erythrocytes (Jennings, 1989; Arakawa et al., 2015). Altogether, these results suggested that NBCn1 represents a plausible keratinocyte target for SLS during GAS infection.

Although our bioinformatics analysis indicated that NBCn1 was a candidate for targeting by SLS, it was unclear whether NBCn1 was expressed in our HaCaT cell line under our culture conditions. Previous results from the Human Protein Atlas had indicated that SLC4A7 transcript was present in HaCaT cells, but validation of expression at the protein level in this cell line was lacking (Uhlén et al., 2015). Therefore, we isolated total RNA from HaCaT cells and performed RT-PCR using primers that spanned exon-exon junctions and corresponded to cDNA. RT-PCR generated a product of approximately 105 bp corresponding to SLC4A7 mRNA, demonstrating that this gene was expressed in HaCaT cells under our culture conditions ( Figure S3A ). Expression of NBCn1 at the protein level was evaluated using Western blotting. HaCaT cells were lysed in a hypoosmotic lysis buffer, and membranes were isolated by ultracentrifugation. Probing these membrane samples and cytosolic fractions of cell lysates with an antibody to NBCn1 resulted in a band of around 30 kDa in the membrane fraction ( Figure S3B ). Full-length NBCn1 is around 136 kDa in size (UniProt Q9Y6M7). This size discrepancy could be explained by cleavage of the domain recognized by the antibody during the lysis process. Regardless, when taken together, these results demonstrated that the SLC4A7 gene is expressed in HaCaT cells, along with a polypeptide corresponding to a portion of NBCn1. This further suggested that NBCn1 could be targeted by GAS via SLS during infection.

Our RT-PCR and Western blotting data suggested that NBCn1 was expressed in HaCaT cells ( Figure S3 ). Therefore, we set out to determine whether this protein was targeted by SLS during GAS infection. Due to the broad inhibitory activity of DIDS and the stilbene derivatives discussed previously, we sought a more specific chemical inhibitor of sodium-bicarbonate cotransporters such as NBCn1 for use in GAS infections. Cells were treated with the N-cyanosulfonamide S0859 (whose structure is shown in Figure 2A ), prior to infection with our panel of GAS strains. Unlike DIDS and the stilbene derivatives, S0859 is a much more specific inhibitor of sodium-bicarbonate cotransport, despite being unable to distinguish between individual sodium-bicarbonate cotransporters (Ch’en et al., 2008; Boedtkjer et al., 2012). When cells were treated with DMSO as a vehicle control, nearly 50 percent cytotoxicity was once again observed following infections with the SLS-producing GAS strains, but not in the uninfected or SLS-deficient ΔsagA controls ( Figure 2B ). Pre-treatment of keratinocytes with 100 μM S0859 resulted in a significant, SLS-dependent inhibition of cell death in our keratinocyte line, similar to what was seen with DIDS treatment ( Figure 2B ). In order to confirm that this inhibition was not due to S0859-dependent effects on GAS viability, we performed a growth curve analysis using wild-type GAS in the presence of DMSO or 100 µM S0859. This treatment had no apparent effect on GAS growth, which indicated that the inhibitory effects were due to S0859 treatment mitigating SLS-mediated pathology, and not the growth of the GAS strains in our infection conditions ( Figure 2C ). These results suggested that sodium-bicarbonate transport plays an important role during SLS-mediated cytotoxicity during GAS infections of HaCaT keratinocytes. We also evaluated whether S0859 treatment was capable of preventing SLS-mediated hemolysis in erythrocytes, and we determined that 50 µM S0859 treatment was unable to protect erythrocytes from hemolysis in response to SLS ( Figure 2D ). Interestingly, S0859 treatment alone induced hemolysis at an intermediate level between DMSO-treated cells and wild-type GAS-treated cells, suggesting that this compound induced an osmotic stress that the erythrocytes were partially sensitive to. However, there was no significant difference observed in hemolysis when erythrocytes were treated with DMSO or 50 µM S0859 in the presence of SLS preparations from wild-type GAS. Altogether, this indicated that S0859 treatment was able to prevent SLS-dependent effects in some cell types, but not others.

In addition to causing cytotoxicity during HaCaT infection, SLS is also known to induce multiple cell signaling events prior to cell death. These include the activation of pro-inflammatory signals through the transcription factor NF-κB (Flaherty et al., 2015). Therefore, we hypothesized that S0859 treatment inhibited SLS-mediated NF-κB activation during GAS infections, preventing the inflammatory response and subsequent cytotoxicity. To test this, we performed GAS infections of S0859 pre-treated HaCaT cells on glass coverslips for 4 hours. This shorter time course was used to observe signaling events prior to the widespread increase in SLS-dependent cell death that occurred as the infection approached 6 hours. Following the infection, cells were fixed with paraformaldehyde and NF-κB p65 localization was assessed using immunofluorescence microscopy. NF-κB is a transcription factor that is typically bound to its inhibitor IκBα and localized to the cytoplasm. When activated by a stimulus, IκBα is degraded and NF-κB translocates to the cell nucleus, where it activates the transcription of pro-inflammatory genes (Liu et al., 2017). Therefore, the number of cells with NF-κB localized to the nucleus in a particular field can be used to assess the level of NF-κB activation in response to a stimulus, such as GAS infection.

When cells were pre-treated with the vehicle control DMSO, most cells showed NF-κB primarily in the cytoplasm in the uninfected control ( Figure 3A ) or when infected with the ΔsagA strain ( Figure 3C ). However, when HaCaT cells were infected with the SLS-producing wild-type ( Figure 3B ) and ΔsagA + sagA ( Figure 3D ) strains, the majority of cells experienced robust NF-κB localization to the cell nucleus. This aligned with previous results, indicating that SLS induced NF-κB activation in epithelial cells during GAS infection. When cells were pre-treated with S0859, minimal NF-κB activation was observed in the uninfected ( Figure 3E ) and ΔsagA ( Figure 3G ) conditions, similar to the DMSO-treated conditions. However, we also observed reduced NF-κB activation in S0859-treated cells infected with the wild-type ( Figure 3F ) and ΔsagA + sagA ( Figure 3H ) GAS strains. These effects were quantified by counting the number of cells with NF-κB localized to the nucleus relative to the total number of cells in 10 fields per condition, and this analysis demonstrated that treatment with S0859 significantly inhibited the ability of SLS to activate NF-κB ( Figure 3I ). Taken together, these results demonstrated that targeted inhibition of sodium-bicarbonate cotransport with S0859 reduces the ability of SLS to promote inflammation during GAS infection of epithelial cells.

NBCn1 and other SLC4 proteins that are involved in sodium-bicarbonate cotransport are often associated with net acid loading or extrusion, which is accomplished by moving bicarbonate ions across the cell membrane (Boron et al., 2009; Romero et al., 2013; Thornell and Bevensee, 2015). By extension, these proteins play a major role in the maintenance of internal pH (pH_i). Given that chemical inhibition of sodium-bicarbonate cotransport with S0859 was able to reduce the pathological effects of SLS ( Figures 2 , 3 ), we hypothesized that SLS would disrupt the ability of HaCaT cells to maintain their pH_i during infection. To test this, we loaded HaCaT cells with the pH-sensitive, fluorescent dye pHrodo Red AM, infected cells with live GAS, and monitored the fluorescent response within the cells. The high photostability and retention of pHrodo Red AM within cells makes ratiometric measurements unnecessary, and because the fluorescent intensity of pHrodo Red AM increases as pH_i decreases, a single fluorescence reading can be directly correlated to pH_i levels.

Therefore, we subjected pHrodo Red AM-loaded HaCaT cells to live GAS infections for 8 hours, and monitored fluorescence intensity over time using real-time live imaging. Infections were carried out at an MOI of 0.01 to prevent keratinocytes from being rapidly overrun by bacteria, and images were taken every 10 minutes over the course of the infection. Exposure time for the pHrodo channel was 10 ms across all conditions. Cells infected with either the wild-type or complemented ΔsagA + sagA strain of GAS experienced a sharp increase in fluorescence intensity over time, with some cell boundaries eventually becoming indistinguishable due to the strength of the fluorescence ( Videos S1A, S3A ). In contrast, cells infected with the SLS-deficient ΔsagA strain experienced a much more gradual increase in fluorescence intensity over time ( Video S2A ). This was clearly seen in frames taken every 120 minutes from representative videos for each infection condition, as shown in Figure S4 . Images were also taken in the DIC channel to monitor bacterial growth over the course of the infection, and these videos indicated that bacterial growth was comparable between all conditions ( Videos S1B–S3B ). Eight cells per condition were tracked over the course of the 8 hours to observe changes in corrected total cell fluorescence over time, but the variability between cells within each condition resulted in large error that obscured any differences between the strains (data not shown). Regardless, our data demonstrated that infection of HaCaT cells with GAS strains that produce SLS experience a more dramatic intracellular acidification than those that are infected with GAS strains that do not produce SLS.

Having shown that SLS exposure resulted in a decrease in pH_i, we hypothesized that S0859 exhibited its protective effects on keratinocyte cytotoxicity by preventing or disrupting SLS from inducing this intracellular acidification. To test this, we performed real-time live imaging experiments with HaCaT cells that were pre-treated with 100 µM S0859 prior to infection with wild-type GAS. Interestingly, these pre-treated cells also experienced a rapid increase in fluorescence intensity over the 8 hour infection period ( Video S4A ). The DIC channel indicated that bacterial growth was similar to what we observed in the videos of cells not treated with S0859 ( Video S4B ). These results indicated that SLS exposure in either the presence or absence of S0859 treatment caused a decrease in keratinocyte pH_i, and suggested that S0859 inhibits SLS-mediated cytotoxicity through a mechanism that does not involve preventing an intracellular acidification event. Furthermore, it is likely that S0859 treatment alone would result in intracellular acidification by disrupting normal sodium-bicarbonate cotransport. Altogether, these results from monitoring pH_i changes demonstrated that epithelial cells experience strong intracellular acidification in response to SLS.

Having established that sodium-bicarbonate cotransport played a role in SLS-mediated GAS pathogenesis in vitro, we hypothesized that S0859 treatment would mitigate SLS-dependent effects in vivo. To test this, we used C57BL/6 mice expressing the human plasminogen transgene as a model system for GAS infection, which has been described previously (Sun et al., 2004; Higashi et al., 2016). Mice were injected with 10⁷ CFU of either wild-type (± 100 µM S0859), ΔsagA, or ΔsagA + sagA GAS, and were monitored for 3 days. Mice infected with wild-type GAS in both the presence and absence of S0859 treatment developed skin lesions within 24 hours of infection, while wounds in the ΔsagA, and ΔsagA + sagA conditions progressed more slowly ( Figure S5 ). By 72 hours post-infection, there was no significant quantitative difference in wound size between wild-type-infected mice treated with the vehicle control or 100 µM S0859. S0859 has been described as a reversible inhibitor (Ch’en et al., 2008), so we hypothesized that additional injections of a higher dose of S0859 would be able to produce a treatment effect. We injected mice with wild-type GAS and either 200 µM S0859 or 0.7% saline as a vehicle control at the start of the infection, and then provided additional injections of either 200 µM S0859 or saline at 4 and 8 hours post-infection. Both groups of mice developed wounds within 24 hours, and there was no apparent difference in the size of these wounds ( Figures 4A, C ) or bacterial CFU recovered from the wounds between groups ( Figure 4B ). This indicated that S0859 was unable to inhibit SLS-mediated GAS virulence in vivo, which could be attributed to either the reversible nature of the inhibitor or its stability in vivo. In addition, the ability of SLS to still cause pathology in this model despite S0859 treatment indicates that SLS is still likely contributing to lesion development by exhibiting cytolytic effects on host cells other than keratinocytes, such as erythrocytes. This supports the hypothesis that SLS has multiple targets in different host cell types, because the in vivo system likely has multiple host cells and components that are targeted by SLS during GAS infection.