The major role of sarA in limiting Staphylococcus aureus extracellular protease production is correlated with decreased virulence in diverse clinical isolates in osteomyelitis

We previously demonstrated that MgrA, SarA, SarR, SarS, SarZ, and Rot bind at least three of the four promoters associated with genes encoding primary extracellular proteases in Staphylococcus aureus. We also showed that mutation of sarA results in a greater increase in protease production, and decrease in biofilm formation, than mutation of the loci encoding any of these other proteins. However, these conclusions were based on in vitro studies. Thus, the goal of the experiments reported here was to determine the relative impact of the regulatory loci encoding these proteins in vivo. To this end, we compared the virulence of mgrA, sarA, sarR, sarS, sarZ, and rot mutants in a murine osteomyelitis model. Mutants were generated in the methicillin-resistant USA300 strain LAC and the methicillin-sensitive USA200 strain UAMS-1. As assessed based on an overall osteomyelitis pathology score derived from the incidence of bone fracture, bacterial burdens in the bone, cortical bone destruction, and reactive bone formation, mutation of mgrA and rot limited virulence to a statistically significant extent in UAMS-1, but not in LAC. In contrast, the sarA mutant exhibited reduced virulence in both strains. This illustrates the importance of considering diverse clinical isolates when evaluating the impact of regulatory mutations on virulence. The reduced virulence of the sarA mutant was correlated with reduced cytotoxicity for osteoblasts and osteoclasts, reduced biofilm formation, and reduced sensitivity to the antimicrobial peptide indolicidin, all of which were directly attributable to increased protease production in both LAC and UAMS-1. This suggests that these in vitro phenotypes, either alone or in combination with each other, may be useful in prioritizing additional mutants for in vivo evaluation. Most importantly, they illustrate the significance of limiting protease production in vivo in S. aureus, and confirm that SarA plays the primary role in this regard. Author Summary Staphylococcus aureus causes a diverse array of infections due to its ability to produce an arsenal of virulence factors. Among these are extracellular proteases, which serve several purposes on behalf of the bacterium. However, it has become increasingly apparent that it is also critical to limit the production of these proteases to prevent them from compromising the S. aureus virulence factor repertoire. Many regulatory loci have been implicated in this respect, but it is difficult to draw relative conclusions because few reports have made direct comparisons, and fewer still have done so in vivo. We addressed this by assessing the impact on virulence of six regulatory loci previously implicated in protease production. We did this in the clinical context of osteomyelitis using mutants generated in two divergent clinical isolates. Our results confirm significant strain-dependent differences, reinforcing the importance of considering such diverse clinical isolates when evaluating targets for potential therapeutic intervention. In this respect, only mutation of sarA attenuated virulence in both strains. This illustrates the importance of limiting protease production as a means of post-translational regulatory control in S. aureus and confirms that sarA plays a predominant role in this regard.

Impact of regulatory loci on osteomyelitis-associated 153 phenotypes in LAC.

154
The results observed with LAC and its regulatory mutants in our osteomyelitis model were 155 directly reflected in studies assessing the impact of regulatory mutants on osteoblast and 156 osteoclast cytotoxicity. Specifically, conditioned medium (CM) from LAC was cytotoxic for both 157 cell types, and the only mutation that significantly reduced this cytotoxicity was sarA (Fig 2A and  158 B). Additionally, cytotoxicity was fully restored in the LAC sarA mutant by eliminating the 159 production of extracellular proteases (Fig 2A and B). Mutation of sarA also limited biofilm 160 formation in LAC to a statistically significant degree, and this was also reversed by eliminating 161 the ability of the sarA mutant to produce extracellular proteases ( Fig 2C). Mutation of rot also 162 reduced the capacity of LAC to form a biofilm, and this effect was also reversed by eliminating 163 the ability of the rot mutant to produce extracellular proteases ( Fig 2C). Growth in the presence 164 of indolicidin, which we used as a representative antimicrobial peptide (AMP), was significantly 165 increased in the sarA mutant in a protease dependent manner ( Fig 2D). No other regulatory 166 mutants had a significant increase in growth in the presence of indolicidin by comparison to 167 LAC, but mutation of sarZ resulted in a significant reduction in growth ( Fig 2D). Survival in whole 168 human blood was increased to a significant extent in LAC sarA mutant ( Fig 2E). However, 169 eliminating the production of extracellular proteases in the sarA mutant resulted in only a slight 170 the sarA mutant ( Fig 2E). was mixed in a 1:1 ratio with the appropriate cell culture medium for cytotoxicity assays. C) 178 Biofilm formation was assessed using a microtiter plate assay. The impact of mutating rot on 179 biofilm formation was statistically different from that of mutating sarA (p = 0.0214). D) For each 180 strain, 1x10 6 cfu was inoculated into TSB with or without 10 μg/mL indolicidin and incubated 181 overnight with shaking. Growth was measured by OD600 relative to a DMSO control for each 182 mutant. E) 1x10 5 cfu of bacterial cells from exponential phase cultures were mixed with 1.0 ml of 183 whole human blood. A sample was taken immediately after mixing and after a 3 hr incubation. 184 Percent survival of each mutant was calculated and standardized relative to the results 185 observed with LAC. The difference between the results observed with the sarA mutant and its 186 protease-deficient derivative were not statistically significant. In all cases, statistical significance 187 was assessed by one-way ANOVA. Numbers indicate p-values by comparison to the results 188 observed with the LAC parent strain. 189 Impact of regulatory loci on virulence in UAMS-1. definitive as those observed in LAC. For instance, no broken bones were observed in either of 192 two independent experiments in mice infected with the sarA, mgrA, or sarR mutants ( Fig 3A). 193 Moreover, while clear downward trends were observed with UAMS-1 sarA and mgrA mutants, 194 no statistically significant differences were observed in cortical bone destruction (Fig 3B). 195 Mutation of sarA, mgrA or rot did result in a statistically significant reduction in new bone 196 formation (Fig 3C), and mutation of mgrA or rot resulted in significantly decreased bacterial 197 burdens while mutation of sarA resulted in a downwards trend that did not reach statistical 198 significance (p-value 0.0526) ( Fig 3D). When all of the in vivo data was combined to generate 199 an overall osteomyelitis score, mutation of sarA, mgrA, or rot were all found to result in a 200 statistically significant reduction in virulence ( Fig 3E). Impact of regulatory loci on osteomyelitis-associated 216 phenotypes in UAMS-1.

217
As was observed in LAC, only mutation of sarA resulted in a significant increase in 218 cytotoxicity for osteoblasts and osteoclasts ( Fig 4A and B), and this was directly correlated with 219 the increased production of extracellular proteases (Fig 4A and B). Similarly, mutation of sarA 220 resulted in decreased biofilm formation ( Fig 4C) and increased growth in the presence of 221 indolicidin (Fig 4D), both of which were defined by the increased production of extracellular 222 proteases (Fig 4C and D). As in LAC, survival in whole human blood was also increased in a 223 UAMS-1 sarA mutant, and this phenotype was not protease dependent (Fig. 4E). In contrast to 224 LAC, mutation of rot did not result in a statistically significant decrease in biofilm formation (Fig  225   4C) and mutation of sarZ did not result in significantly decreased growth in the presence of 226 indolicidin ( Fig 4D). sterile bacterial culture media (negative control), or ethanol (positive control) was mixed in a 1:1 232 ratio with the appropriate cell culture medium for cytotoxicity assays. C) Biofilm formation was 233 assessed using a microtiter plate assay. D) For each strain, 1x10 6 cfu was inoculated into TSB 234 with or without 10 μg/mL indolicidin and incubated overnight with shaking. Growth was 235 measured by OD600 relative to a DMSO control for each mutant. E)1x10 5 cfu of bacterial cells 236 from exponential phase cultures were mixed with 1.0 ml of whole human blood. A sample was 237 taken immediately after mixing and after a 3 hr incubation. Percent survival of each mutant was 238 calculated and standardized relative to the results observed with UAMS-1. The difference 239 between the results observed with the sarA mutant and its protease-deficient derivative were 240 not statistically significant. In all cases, statistical significance was assessed by one-way 241 ANOVA. Numbers indicate p-values by comparison to the results observed with the UAMS-1 242 parent strain. 243

244
Staphylococcus aureus is arguably the most diverse of all bacterial pathogens given its 245 ability to cause such a wide array of infections [43]. Among the many virulence factors which 246 mediate this diversity are the extracellular proteases aureolysin, SspA/SspB, ScpA, or SplA-F, 247 which have been shown to promote nutrient acquisition, tissue invasion, and evasion of host 248 defenses [28,29,[44][45][46]. However, recent reports have demonstrated that eliminating protease 249 production results in increased virulence, while mutants that exhibit increased protease 250 production also exhibit reduced virulence [22][23][24][25]38,47]. Neither of these would be predicted for 251 a classic virulence factor. 252 factor repertoire of S. aureus. Specifically, eliminating protease production increases virulence 254 because it results in the increased abundance of other virulence factors [38,47]. Conversely, 255 increased protease production has been shown to decrease virulence because it results in the 256 decreased abundance of these virulence factors [21,25,26,30,48]. Such results suggest that 257 extracellular proteases also serve a key post-translational regulatory role that must be kept in 258 check to allow the proteases to serve their intended purposes on behalf of the bacterium without 259 compromising the rest of the S. aureus virulence factor repertoire. 260 The importance of this post-translational control is reinforced by the number of regulatory 261 loci that have been implicated in modulation of extracellular protease production 262 [22,26,33,37,40]. However, it is impossible to put the relative role of these regulatory loci into 263 context because few reports have included direct comparisons. It is also not possible from these 264 reports to determine whether the impact of different regulatory loci occurs via a direct or indirect 265 mechanism, particularly given the complexity and highly interactive nature of S. aureus 266 regulatory circuits [49,50]. To begin to address these issues, we used the promoters associated These results suggest that sarA plays a major role in limiting the production of extracellular 275 proteases and that, by doing so, plays a predominant role in mediating post-translational 276 regulation in S. aureus. However, these studies were limited to in vitro experiments and 277 therefore do not necessarily reflect in vivo significance, particularly in a unique 278 relevance of the genes encoding the regulatory proteins captured in our earlier experiments 280 (mgrA, rot, sarA, sarR, sarS, and sarZ). We chose to use a murine osteomyelitis model based 281 on our specific interest in overcoming the therapeutic recalcitrance of orthopaedic infections and 282 because we had previously shown that increased protease production is correlated with 283 decreased virulence in this model [22,25,48,51]. We also chose to include mutants generated in 284 both the methicillin-resistant USA300 strain LAC and the methicillin-sensitive USA200 strain allowed us to determine whether we could identify common elements for potential therapeutic 293

intervention. 294
In LAC, the results of our studies were very clear in that mutation of sarA resulted in a 295 decrease in osteomyelitis virulence as reflected by cortical bone destruction, reactive new bone 296 formation, bacterial burdens in the femur, and overall osteomyelitis scores, while none of the 297 other regulatory mutants examined impacted any of these phenotypes to a statistically 298 significant degree. This result could be correlated with cytotoxicity for osteoblasts and 299 osteoclasts, which was reduced in CM from a LAC sarA mutant but was not altered with CM 300 from any other regulatory mutant. Moreover, the reduced cytotoxicity of CM from the sarA 301 mutant was a direct function of the increased production of extracellular proteases as evidenced 302 by the fact cytotoxicity was restored by eliminating the ability of the sarA mutant to produce 303 aureolysin, ScpA, SspA, SspB, and the spl-encoded proteases (SplA-F). Biofilm formation was 304 sarA mutant. Mutation of rot reduced biofilm formation in LAC in a protease-dependent manner, 306 although to a lesser extent than mutating sarA. 307 Sensitivity to the antimicrobial peptide indolicidin was also decreased in the LAC sarA 308 mutant in a protease-dependent manner. In contrast, sensitivity to indolicidin was increased in a 309 LAC sarZ mutant. The reasons for this remain to be determined, but it has been shown that 310 mutation of sarZ results in the increased transcription of sarA and decreased transcription of the 311 gene encoding sspA [60]. Interestingly, survival in whole human blood was also increased to a 312 statistically significant extent in the LAC sarA mutant, and this was not the case with any other 313 regulatory mutant. However, eliminating protease production did not reverse this phenotype to a 314 statistically significant extent, possibly suggesting that other loci regulated by sarA are involved. 315 Overall, these studies are consistent with the conclusions that S. aureus must carefully limit the 316 production of extracellular proteases and that SarA plays a direct and predominant role in this 317

regard. 318
The results observed with UAMS-1 were consistent but less definitive than those observed 319 with LAC. Specifically, mutation of sarA in UAMS-1 limited virulence in our osteomyelitis model, 320 but not to the same extent observed with a LAC sarA mutant. Mutation of sarA had the same 321 effect in both strains on cytotoxicity, biofilm formation, sensitivity to indolicidin, and survival in 322 whole human blood, and, as with a LAC sarA mutant, eliminating protease production reversed 323 all of these phenotypes except survival in blood. Thus, the results we report confirm that sarA 324 plays an important role in vivo in limiting protease production, and that it does so in diverse 325 clinical isolates. However, significant strain-dependent differences were observed with respect 326 to other regulatory loci. 327 For example, mutation of sarZ had no impact on indolicidin sensitivity in UAMS-1. Similarly, Gimza et al [37] concluded that mgrA is one of the primary regulators of protease 348 production. It affects at least 9 other regulatory loci (agr, arIRS, atIR, sarR, sarS, sarV, sarX, 349 sarZ, sigB) and can be regulated by at least 7 others (agr, arIRS, mntR, rex, sarV, sarZ, and Thus, the reason(s) mutation of mgrA and rot limit the virulence of UAMS-1 in osteomyelitis 355 remain to be determined, but, irrespective of the mechanism(s) involved, these results do not 356 20 contradict the conclusion that sarA is the only regulatory locus among those we examined that 357 is consistently associated with reduced virulence in diverse clinical isolates of S. aureus. These 358 results emphasize the importance of in vivo comparative studies to determine the relative 359 impact of regulatory loci in the context of both infection and the diversity among clinical isolates 360 of S. aureus. They also suggest that sarA is one of if not the most promising regulatory loci for 361 the development of anti-virulence therapies targeting osteomyelitis owing to its role in limiting 362 protease production as an important means of post-translational regulation of the S. aureus 363 virulence factor repertoire. 364 In conclusion, the results we report emphasize the importance of in vivo comparative 365 studies to determine the relative impact of regulatory loci in the context of infection and the 366 diversity among clinical isolates of S. aureus. They also demonstrate that mutation of sarA 367 attenuates the virulence of both LAC and UAMS-1 to a greater and more consistent extent than 368 mutation of any of the other regulatory loci we examined. The results of our cytotoxicity and 369 biofilm studies with both LAC and UAMS-1 are consistent with the hypothesis that this 370 attenuation can be attributed to the increased production of extracellular proteases in sarA 371 mutants, and in fact this has been proven in vivo in our osteomyelitis model using LAC [25]. 372 Studies to confirm that this is also the case in a UAMS-1 sarA mutant and to determine the 373 relative contribution of specific proteases are ongoing. Most importantly, our results point to 374 sarA as the most promising protease regulatory locus for the development of anti-virulence 375 therapies targeting osteomyelitis. 376

377
Bacterial strains and growth conditions. 378 LAC and UAMS-1 regulatory mutants were generated by phage-mediated transduction 379 using mutants available in the NTML as donor strains [23]. Derivatives of each parent strain 380 and specific regulatory mutants with additional mutations in the genes encoding aureolysin, 381 strains were maintained as stocks at -80°C in tryptic soy broth (TSB) supplemented with 25% 383 (v/v) glycerol. Bacteria were recovered from frozen stocks by plating on tryptic soy agar (TSA) 384 containing appropriate antibiotics. Antibiotics were used at the following concentrations: 385 chloramphenicol, 10 µg/ml; kanamycin, 50 µg/ml; neomycin, 50 µg/ml; erythromycin, 10 µg/ml; 386 spectinomycin, 1 mg/ml; and tetracycline, 5 µg/ml. Prior to in vivo analysis, each strain was 387 grown overnight at 37°C in TSB with shaking and without antibiotic selection, washed 3 times 388 with sterile phosphate-buffered saline (PBS), and then resuspended in PBS at a density of 5 X 389 10 8 colony-forming units (cfu) per ml. The concentration of each strain was confirmed by plating 390 serial dilutions on TSA with and without appropriate antibiotic selection. 2 µl of this suspension 391 (1 X 10 6 cfu) was then used to infect mice. 392 Murine osteomyelitis model. with NIH guidelines, the Animal Welfare Act, and United States federal law [25,51]. Briefly, 6-8 397 week-old C57BL/6 mice were anesthetized and the femur exposed by making an incision in the 398 right hind limb. A unicortical defect was created in the middle of the exposed femur. 1 × 10 6 399 bacterial cells prepared as described above were then injected into the medullary canal in a 400 total volume of 2 μl. Muscle and skin were sutured, and the infection allowed to proceed for 14 401 days. Mice were humanely euthanized and the infected femurs recovered. After removing soft 402 tissues, femurs were frozen at -80°C before imaging by microcomputed tomography (μCT). 403 After imaging, femurs were homogenized and bacterial burdens determined as detailed below. 404 At least two independent experiments with 5 mice per experimental group were done with LAC, 405

UAMS-1, and their isogenic mutants. 406
In the first set of experiments done with each set of strains, image acquisition was done 408 using a Skyscan 1174 X-ray Microtomograph (Bruker, Kontich, Belgium) with an isotropic voxel 409 size of 6.7 μm, an X-ray voltage of 50 kV (800 μA) and a 0.25 mm aluminum filter [22,25]. 410 Reconstruction was carried out using the Skyscan Nrecon software. The reconstructed cross-411 sectional slices were processed using the Skyscan CT-analyzer software as previously 412 described to delineate regions of interest (ROIs) where reactive new bone (callus) was isolated 413 from cortical bone [22,25]. The ROIs were used to calculate the volume of cortical bone, and the 414 amount of cortical bone destruction was estimated by subtracting the value obtained from each 415 bone from the average obtained from sham operated bones inoculated with PBS. New bone 416 formation was quantified using the subtractive ROI function on the previously delineated cortical 417 bone-including ROI images and calculating the bone volume included in the newly defined ROI. 418 In the second set of experiments, image acquisition was done using a Skyscan 1275 X-ray 419 Microtomograph (Bruker, Kontich, Belgium) with an isotropic voxel size of 6.8 μm and an X-ray 420 voltage of 40 kV (100 μA). Reconstruction was carried out using the Skyscan Nrecon software. 421 The reconstructed cross-sectional slices were processed using the Skyscan CT-analyzer 422 software to perform a semiautomated protocol to create preliminary ROIs of only cortical bone. 423 The semi-automated protocol was as follows: global thresholding (low = 90; high = 255), round 424 closing in 3D space pixel size 4, round opening in 3D space, pixel size 1, round closing in 3D 425 space, pixel size 8, and round dilation in 3D space pixel size 3. The resulting images were 426 loaded as ROI and corrected by drawing inclusive or exclusive contours on the periosteal 427 surface to keep only the cortical bone. Using these defined ROIs, the volume of cortical bone 428 was calculated using a threshold of 70-255, and the amount of cortical bone destruction 429 estimated by subtracting the value obtained from each bone from the average obtained from 430 sham operated bones inoculated with sterile PBS. New bone formation was quantified using the 431 the bone volume included in the newly defined ROI using a threshold of 45-135. 433 In both sets of experiments, statistical analysis was done by one-way ANOVA with 434 Dunnett's correction. Comparisons were made with all mutants relative to the appropriate parent 435 strain. A p-value ≤0.05 was considered statistically significant. 436 Bacterial burdens in the femur. Osteomyelitis score.

453
In some cases, quantitative µCT analysis was not possible because the femur was broken. 454 We addressed this based on the premise that fracture was indicative of pathology by developing 455 an osteomyelitis (OM) score for each experimental animal. Therefore, in the absence of 456 24 fracture, the score formula was based on the sum of the amount of cortical bone destruction 457 (mm 3 ) + the amount of reactive bone formation (mm 3 ) + the log10 of the cfu per femur. In those 458 mice in which the bone was fractured, the numbers used for cortical bone destruction and 459 reactive bone formation were derived by adding one standard deviation to the highest scores 460 observed with an intact bone from the same experimental group. Statistical analysis was done 461 by comparing OM scores for individual mice in each experimental group by one-way ANOVA 462 with Dunnett's correction. A p-value ≤0.05 was considered statistically significant. 463 Cytotoxicity for mammalian cells. Sensitivity to indolicidin.

484
Sensitivity to the antimicrobial peptide indolicidin was assessed as previously described 485 [47,78] with modification. Specifically, each strain was grown overnight in TSB at 37°C with 486 shaking. Cultures were standardized in TSB and 1 x 10 6 cfu added to the wells of a 96-well 487 microtiter plate containing 50 μl of 2-fold concentrated TSB, 30 μl of sterile water, and 20 μl of 488 50 μg/ml indolicidin in DMSO. DMSO without indolicidin was used as a control. After overnight 489 incubation at 37°C with shaking, optical density at 600 nm (OD600) was determined using a 490 microtiter plate reader and the percent growth calculated as the OD600 in the presence of 491 indolicidin divided by the OD600 without indolicidin X 100. Results are reported as the average of 492 3 biological replicates, each of which included 3 experimental replicates. 493 Survival in whole human blood.

494
Survival in whole human blood was performed as previously described [47,78] with minor 495 modification. Briefly, strains were grown overnight in TSB at 37°C with shaking, diluted 1:100 in 496 4 ml TSB, and incubated for an additional 3 hours. Cells from 1 ml of each culture were 497 harvested by centrifugation, washed twice in PBS, and standardized to an equivalent optical 498 density in PBS. 1 x 10 6 cfu was then added to 1 ml of whole human blood (BioIVT). An aliquot 499 was immediately removed, serially diluted, and plated to verify the inoculum. The mixture was 500 then incubated at 37° with shaking for 3 h, after which another aliquot was removed, diluted, 501 and plated. Relative percent survival was determined by finding the percent survival of each 502 mutant [(cfu at 3 hours divided by cfu at 0 hours] X 100) and dividing by the percent survival of 503 the appropriate parent strain in each experiment. Experiments were done as at least 3 biological 504 replicates, each with 3 experimental replicates. 505 We would like to acknowledge Horace "Trey" J. Spencer in the University of Arkansas for 507 Evolving concepts in bone infection: redefining "biofilm", "acute vs. chronic osteomyelitis", 530 "the immune proteome" and "local antibiotic therapy." Bone Res. 2019;7: 20. Free Proteomic Approach to Characterize Protease-Dependent and -591