Contribution of metal transporters of the ABC, ZIP, and NRAMP families to manganese uptake and infective endocarditis virulence in Streptococcus sanguinis

Streptococcus sanguinis is an important cause of infective endocarditis. In strain SK36, the ABC-family manganese transporter, SsaACB, is essential for virulence. We have now identified a ZIP-family protein, TmpA, as a secondary manganese transporter. A tmpA mutant had no phenotype, but a ΔssaACB ΔtmpA mutant was far more attenuated for serum growth and somewhat more attenuated for virulence in a rabbit model than its ΔssaACB parent. The growth of both mutants was restored by supplemental manganese, but the ΔssaACB ΔtmpA mutant required twenty-fold more and accumulated less. Although ZIP-family proteins are known for zinc and iron transport, TmpA-mediated transport of either metal was minimal. In contrast to ssaACB and tmpA, which appear ubiquitous in S. sanguinis, a mntH gene encoding an NRAMP-family transporter has been identified in relatively few strains, including VMC66. As in SK36, deletion of ssaACB greatly diminished VMC66 endocarditis virulence and serum growth, and deletion of tmpA from this mutant diminished virulence further. Virulence was not significantly altered by deletion of mntH from either VMC66 or its ΔssaACB mutant. This and the accompanying paper together suggest that SsaACB is of primary importance for endocarditis virulence while secondary transporters TmpA and MntH contribute to growth under differing conditions.

The Virginia Commonwealth University Institutional Animal Care and Use Committee approved 28 all animal procedures (IACUC Protocol #AM10030), ensuring compliance with PHS Policy on 29 Humane Care and Use of Laboratory Animals, The Guide for the Care and Use of Laboratory 30 Animals, and all other applicable regulations. 31 Introduction 53 Infective endocarditis (IE) is a disease caused by microorganisms entering the bloodstream and 54 colonizing damaged heart valves, leading to potentially fatal complications such as congestive 55 heart failure, aneurysm, and stroke (Bashore et al., 2006). Many recent studies suggest that IE 56 incidence is rising (Quan et al., 2020), and morality rates of 12-30% are common (Jamil et al.,57 2019, Ly et al., 2020). Currently, prevention is limited to prophylactic antibiotics before dental 58 procedures (Wilson et al., 2021). The economic burden, potential for side effects, and 59 questionable efficacy (Dayer & Thornhill, 2018, Thornhill et al., 2018, Quan et al., 2020 of this 60 practice, as well as the increasing prevalence of antibiotic resistance (Dodds, 2017) all suggest 61 the need for new approaches to prevention. In addition, the ability of oral bacteria to enter the 62 bloodstream through any opening in the oral mucosa, which may occur during routine hygiene 63 practices or mastication (Wray et al., 2008, Wilson et al., 2007, explains why prophylaxis given 64 before dental procedures can never prevent all IE. Thus, it would be desirable to identify an 65 inhibitor that is specific to endocarditis pathogens and which could be taken on a daily basis 66 without disturbing the microbiome or selecting for resistance. 67 Streptococcus sanguinis is one of the most common oral bacteria to be isolated from IE patients 68 (Di Filippo et al., 2006). It is typically considered a commensal in the oral cavity due to an 69 antagonistic relationship with the caries pathogen Streptococcus mutans (Kreth et al., 2005). The has been confirmed in at least two protein families: ATP-binding cassettes (ABC) and natural 74 resistance-associated macrophage proteins (NRAMP) (Waters, 2020). A knockout mutant of the 75 lipoprotein component of an ABC manganese transporter in S. sanguinis, SsaB, is deficient in 76 manganese transport as well as in aerobic serum growth (Crump et al., 2014). This growth defect 77 was rescued by the addition of only 2 µM Mn 2+ , indicating that manganese is able to enter S. 78 sanguinis cells despite the absence of the primary transporter. The genome annotation of S. 79 sanguinis SK36 (Xu et al., 2007) did not indicate the presence of any canonical manganese 80 transporters, which led us to examine other metal transport protein families. NRAMP-family 81 proteins (Nevo & Nelson, 2006) are encoded by at least eight S. sanguinis strains but not by 82 SK36. These proteins, often named MntH, have been found to contribute to manganese uptake 83 and acid tolerance in other streptococci (Shabayek et al., 2016, Kajfasz et al., 2020. 84 The family of ZRT-, IRT-like proteins (ZIP) is well-known for its role in the transport of zinc 85 (Zn), iron (Fe), or other metals across cellular membranes (Eide, 2004). The ZIP family takes its 86 name from the first identified members: zinc regulated transporters (ZRT1 and ZRT2) found in 87 Saccharomyces cerevisiae (Zhao & Eide, 1996a, Zhao & Eide, 1996b) and iron regulated 88 transporter (IRT1) from Arabidopsis thaliana (Eide et al., 1996). Since these initial discoveries, 89 ZIP-family proteins have been identified in organisms of various phyla, including 14 in humans 90 (Jeong & Eide, 2013). While ZIP-family proteins principally transport zinc or iron, two human 91 versions, hZIP8 (Park et al., 2015, Fujishiro & Himeno, 2019) and hZIP14 (Scheiber et al., 2019, 92 Aydemir et al., 2017), as well as BmtA from Borreliella burgdorferi (Ouyang et al., 2009, 93 Ramsey et al., 2017) primarily transport manganese. Bacterial ZIP proteins fall into the GufA 94 subfamily (Gaither & Eide, 2001), which also contains mammalian members such as hZIP11 95 (Dempski, 2012, Yu et al., 2013. The first bacterial ZIP protein, ZupT, was identified in 96 Escherichia coli (Grass et al., 2002). This initial study proved that it played a role in zinc uptake, 97 and further investigation determined that other metal cations could also be transported by ZupT,98 albeit with lower affinity (Grass et al., 2005, Taudte & Grass, 2010. Many bacterial species 99 contain putative ZIP-family proteins, but few have been characterized for metal affinity and 100 contribution to growth and virulence. 101 Here we confirm that ZIP-family proteins should be considered an additional family of bacterial 102 manganese importers. We report that a ZIP-family protein in S. sanguinis contributes to 103 manganese uptake in an Δ ssaACB mutant, deleted for the ABC-family manganese transporter, 104 and we establish its contribution to IE virulence. We extended this study to additional S. 105 sanguinis strains, including those that encode MntH homologs. To our knowledge, we have 106 performed the most extensive analysis of the role of distinct protein families in manganese 107 uptake and virulence that has ever been performed in any Streptococcus. 108

109
Identification of a ZIP-family protein 110 Given the low concentration of manganese required to improve the growth of the differences in growth ( Figure 1A). When the O 2 concentration was reduced to 6% or 1%, growth 144 of the double mutant at 24 h was significantly less than the Δ ssaACB parent strain ( Figure 1A). 145 In contrast, growth of the single mutant was not statistically different from WT under any of the 146 tested O 2 concentrations ( Figure 1A). Additional studies in BHI showed that again, the of the mutant strains in serum, used here as a biologically relevant manganese-deficient medium. 154 Significant differences in growth of the two strains were observed for every added manganese 155 concentration up to 20 µM ( Figure 2). SsaB was also found to transport iron (Crump et al., 2014) 156 and we found that the growth of the Δ ssaACB mutant was maximized by the addition of 100 Fe 2+ (Figure 2) . Growth of the double mutant was significantly less than the parent in every tested 158 Fe 2+ concentration. Growth of the two mutant strains was not significantly different in any 159 concentration of added Zn 2+ (Figure 2), despite the fact that ZIP-family proteins characterized 160 previously primarily transport zinc or iron (Dempski, 2012). These results suggest that TmpA 161 may contribute to both manganese and iron transport in serum. 162 In another test of complementation with added metals, 24-h growth of the double mutant on 163 Todd-Hewitt + Yeast Extract (THY) plates required added Mn 2+ ( Figure S1).  (Figure 3). This result indicates 172 that overexpression of tmpA leads to an increase in growth. The addition of 10 µM Mn 2+ without 173 IPTG to the complemented strain also improved growth to WT levels. Presumably this is due to 174 leakiness of the Phyper-spank promoter, leading to considerable expression of tmpA even 175 without IPTG present. We have observed this previously with the same inducible promoter and 176 expression site (Rhodes et al., 2014). These results confirm that the phenotype of the double 177 mutant is due to loss of TmpA rather than any unintended mutation and suggests that even low-178 level expression of the tmpA gene at its native location would be sufficient to augment the 179 growth of the Δ ssaACB mutant to the extent indicated in Figure 1.  Additionally, there were no difference in magnesium, manganese, iron, or zinc when a different 192 metal was added ( Figure S2). 193 Since the manganese-dependent phenotype of tmpA was only observed in the  Figure S3A). TPEN is often used to reduce available zinc (Ganguly et al., 2021). An added to the media and found that there were no significant differences in any added Zn 2+ 210 concentration ( Figure S3B). We then measured the cellular metal content of these mutant strains 211 in cBHI ( Figure S3C). There was no significant difference in zinc levels between either Δ tmpA 212 mutant and its respective parent, although the slight decrease in the zinc content of the Δ adcC 213 Δ tmpA strain relative to its parent when 10 µM Zn 2+ was added suggests that TmpA may make a 214 minor contribution to zinc transport under these conditions. 215 Expression of the tmpA gene under various metal and oxygen concentrations 216 Bacterial metal transporters are often negatively regulated by the metals that they transport, in 217 most cases by binding of the metal to a protein that represses transcription of the transporter 218 gene(s) (Johnston et al., 2006, Kehres & Maguire, 2003. Therefore, as another approach to 219 investigate the metal specificity of TmpA, we investigated the effect of various metals on the 220 regulation of its gene. WT and yet to discover a condition that leads to differential expression of tmpA and suspect that it may be 235 constitutively expressed. We attempted to assess production of the TmpA protein in S. sanguinis 236 by western blot but were unsuccessful (data not shown). 237

Contribution of TmpA to virulence in a rabbit model of infective endocarditis 238
Given the severe reduction in endocarditis virulence of the Recovery of the Δ tmpA mutant was not significantly different from WT and both were recovered 246 in significantly higher numbers than the Δ ssaACB strain, which was only recovered in one of six 247 rabbits ( Figure 6A). These results indicate that in a WT background, TmpA is not required for 248 virulence in our model, likely because SsaACB can import sufficient manganese to support 249 growth from the low levels found in blood. 250 We next wanted to assess the contribution of tmpA to virulence in a Δ ssaACB background; 251 however, the exceedingly low recovery of the Δ ssaACB mutant made it unlikely that a further 252 reduction in virulence would be detectable in our model. Therefore, for the next experiment, the 253 WT inoculum level was decreased two-fold while the Δ ssaACB and Δ ssaACB Δ tmpA mutant 254 inocula were increased 10-fold, resulting in a 20-fold difference relative to WT. The Δ ssaACB 255 strain was recovered from all six rabbits but at a significantly lower level than WT ( Figure 6B). 256 The recovery of the Δ ssaACB Δ tmpA mutant was significantly lower than the Δ ssaACB mutant. 257 These results suggest that the loss of TmpA in the Δ ssaACB background resulted in a further 258 decrease in virulence, indicating that it may be playing a secondary role in manganese uptake 259 that is only evident when the primary manganese transporter is absent or inactive. 260 Contribution of specific residues of TmpA to its function 261 To examine the differences between TmpA and other ZIP-family proteins, we aligned the amino 262 acid sequence of TmpA to those of two metal-selective ZIP transporters: ZIPB (zinc) and BmtA 263 (manganese) ( Figure S4). We then used Protter (Omasits et al., 2014) to generate a 2D depiction 264 of the protein within a membrane based on the transmembrane domains (TMDs) predicted from 265 the alignment (Figure 7). All ZIP-family proteins characterized thus far have been integral 266 membrane proteins with eight TMDs and are typically predicted to have both C-and N-termini 267 facing extracellularly (Guerinot, 2000). They usually contain two canonical motifs: (i) a variable 268 length (Hx) n motif in the cytoplasmic loop between TMD III and IV (Eide, 2004) and (ii) a 269 conserved HNxPEG motif in TMD IV (Lin et al., 2010). As with BmtA and ZIPB, TmpA has 270 several histidine residues in the variable loop region between TMDs III and IV but only two 271 follow the (Hx) n pattern. Both protein sequences contain the conserved HNxPEG motif in TMD 272 IV. From the alignment ( Figure S4), we found four putative metal-binding residues -E67, N173, 273 E240, and N251 -that were different from the confirmed metal-binding residues from the crystal BmtA ( Figure S4). We decided to mutate all four of these residues in TmpA to alanines to 282 determine the contribution of each side chain to the function of the protein (Morrision & Weiss,  283 2001). We also mutated these four residues to the corresponding residue from ZIPB to determine 284 whether this would affect the function. Since we could not determine a phenotype for the Δ tmpA 285 single mutant, we made these site-directed mutants (SDM) in the Δ ssaACB background. 286 We then assessed the growth of these mutants in comparison to WT, the Δ ssaACB mutant, and 287 the Δ ssaACB Δ tmpA mutant in rabbit serum at 6% O 2 ( Figure 8A). All E67 and E240 mutants 288 grew similarly to the Δ ssaACB parent strain, indicating that these residues are not essential for 289 TmpA protein function. N251H also grew indistinguishably from the parent strain but N251A 290 grew significantly worse, which suggests that this residue may be important for transport, but 291 that histidine is also capable of performing the same function. This was not unexpected, as 292 histidine is known to coordinate manganese (Martin & Giedroc, 2016) and the metal-binding 293 residue at position 251 varied between TmpA and BmtA ( Figure S4). Both N173 mutants grew 294 poorly, which suggests that this residue is critical for function. 295 Despite our best efforts, we were unable to detect TmpA by western blot (Puccio, 2020). 296 However, we were able to conclude that tagged protein was still being produced, as strains with 297 the Strep-Tag® II tag fused to TmpA at two different sites grew similarly to the parent strain 298 when 5 µM Mn 2+ was added ( Figure S5). The clear exception was the N-terminally tagged 299 version, which suggests that the N-terminus may be important either for function or for 300 localization to the membrane. Additionally, we attempted to overexpress the TmpA protein in E. 301 coli strains optimized for heterologous membrane protein expression. While we were able to 302 confirm the presence of TmpA by western blot (Puccio, 2020), we were unsuccessful in 303 obtaining sufficient protein quantities to perform cell-free transport assays in liposomes, which 304 would have been ideal for measuring the contribution of each residue to transport. 305 We thus considered two extreme explanations for the failure of the N173A, N173D, and N251A 306 mutants to display growth that was significantly better than the Δ ssaACB Δ tmpA mutant in 307 Figure 8AFigure 8A: (i) the N173 and N251 mutations interfered with the secretion, localization, 308 or stability of TmpA; or (ii) the SDM were normal with regard to these properties, and their lack 309 of detectable activity was due to the importance of these residues for manganese transport. To 310 distinguish between these two possibilities, the three mutants that grew poorly in serum in 6% O 2 311 were assessed for growth after addition of 5 µM Mn 2+ ( Figure 8B). When Mn 2+ was added, the 312 Δ ssaACB parent grew to WT-like levels, whereas the Δ ssaACB Δ tmpA mutant's growth was still 313 significantly lower than WT. Each of the mutants grew to a level that was lower than the 314 Δ ssaACB parent but higher than the Δ ssaACB Δ tmpA mutant. The results for all three SDM are 315 in agreement with the second model, in which the lack of activity in unamended serum is due to 316 reduced metal transport rather than loss of the protein. 317

Model of TmpA based on the ZIPB crystal structure 318
To determine whether the two N173 mutations would be expected to severely affect metal 319 transport function, we modeled TmpA using the crystal structure of ZIPB as a template ( Figure  320 S6A). The full-length proteins share 40% identity, while the crystal structure of ZIPB (PBD: 321 5TSA) shares 34% identity with TmpA due to lack of structural data for several loops. 322 Additionally, TMD III in TmpA was shorter than that of ZIPB, and thus was depicted missing 323 one of the helical loops in the model ( Figure S6A). The reason for this short TMD is unclear, as 324 the length of all other TMDs appear to match well. 325 We then modelled the N173D mutation in TmpA ( Figure S6B). The other residues that constitute 326 the putative M2 binding site moved to accommodate the negative charge of the aspartic acid, 327 which resulted in a change in the size and shape of the M2 site. The change in shape and size is, 328 of course, speculative, as proteins in vivo are flexible and may be able to accommodate changes 329 such as these. Nevertheless, the predicted structural changes, in conjunction with the obvious 330 change in charge, could well explain the severity of this mutation's effect on metal transport and 331 growth observed in Figure 8. 332 The position of the proteins within the cellular membrane was then predicted using Orientation 333 of Proteins in Membranes (https://opm.phar.umich.edu/). As described in Zhang et al. (2017), the 334 protein has been crystallized with a tilt ( Figure S7A). The model of TmpA fits well in the 335 predicted membrane with this tilt ( Figure S7B). 336 each strain. These mutants and their parent strains were then assessed for serum growth at 6% O 2 349 ( Figure 9). As with SK36 ( Figure 1A) and SK408 versions grew to significantly lower densities than their parent strains. However, it is 353 apparent that in SK678 and VMC66, the ssaACB deletion produced a greater defect on growth 354 than in the other backgrounds ( Figure 9C-D). 355

Evaluation of serum growth of manganese-transporter mutants in additional
To determine if the poor growth of the Δ ssaACB mutant performed so poorly in the 6% O 2 serum growth study ( Figure 9D). To 367 determine whether these NRAMP proteins may contribute to manganese uptake and endocarditis 368 virulence in S. sanguinis, knockout mutants were generated in the VMC66 WT and Δ ssaACB 369 strains and serum growth was assessed. At 6% O 2 , the Δ mntH strain grew to a significantly lower 370 level than the WT parent but higher than the Δ ssaACB mutant ( Figure 10A). Both double mutant 371 strains grew similarly to their Δ ssaACB parent ( Figure 10A). 372 These results indicate that MntH can contribute to growth even in cells possessing the other two 373 transporters, but its contribution is much less than that of SsaACB. As seen in Figure 9, the 374 drastic growth reduction in the Δ ssaACB mutant masked the contribution of TmpA in this 375 background at 6% O 2 and this could have been true in this experiment for MntH as well. Thus, 376 we decided to assess growth at 1% O 2 ( Figure 10B). When we lowered the oxygen concentration, 377 we once again observed that the We then assessed the relative contribution of each secondary transporter to manganese import by 381 measuring cellular metal content by ICP-OES ( Figure 10C). Similar to our previous experiments 382 with SK36 manganese transporter mutants, the low cellular manganese levels that were observed 383 for the VMC66 Δ ssaACB mutants grown in BHI alone made it difficult to evaluate potential 384 differences. To circumvent this issue, we also measured manganese content when 10 µM Mn 2+ 385 was added to the BHI. We found that there were no significant differences between the 386 manganese content of either Δ tmpA mutant and its respective parent strain under either condition 387 ( Figure 10C). The same was true of other tested metals: iron, zinc, and magnesium ( Figure S9). 388 However, when Mn 2+ was added, manganese levels in the Δ ssaACB Δ mntH mutant were 389 significantly lower than in the Δ ssaACB parent, suggesting that MntH contributes more to 390 manganese transport in this background. 391 To determine the relative contribution of each manganese transporter to VMC66 virulence, WT, 392 Δ tmpA, Δ ssaACB, and Δ mntH strains were tested in our rabbit model of IE ( Figure 11A) To test the relative contribution of each secondary transporter to virulence, the double mutants 398 were also tested in our IE rabbit model ( Figure 11B). To ensure sufficient recovery of the 399 Δ ssaACB mutants, we once again increased the inoculum size of each of these mutants and 400 decreased the inoculum size of WT so that the WT inoculum was twenty times less than that of 401 the three mutants. We were able to recover colonies of every strain from each rabbit and we saw 402 a significant difference between the WT and both double mutant strains. Streptococcus vestibularis, all of which appear to lack a TmpA homolog or contain only partial 414 sequences. Additionally, a phylogenetic analysis showed that previously characterized ZIP 415 proteins from other genera of bacteria, such as BmtA and ZIPB, were more closely related to the 416 main group of streptococcal ZIP proteins than were the homologs from S. mutans and 417 Streptococcus ratti ( Figure S10A). The TmpA homolog of Streptococcus sobrinus, a member of 418 the "Mutans" group of streptococci (Nobbs et al., 2009), also clustered with the main group of 419 streptococcal proteins rather than with those from S. mutans and S. ratti. ZupT proteins from E. 420 coli and C. difficile were more similar to the S. mutans protein than to TmpA from S. sanguinis. 421 Furthermore, we included the hu man ZIP-family proteins hZIP11 and hZIP8 with the intention 422 of using them as outgroups. We found that hZIP11 was more closely related to most of the 423 bacterial ZIP-family proteins than were the those of S. mutans and S. ratti, whereas the hZIP8 424 protein functioned as an actual outgroup. These results, along with the observation that our ZIP 425 protein phylogenetic tree looks very different from a 16S rRNA-based tree ( that are closely related to S. mutans ( Figure S11). The S. sanguinis TmpA sequence was included 445 as an outgroup and we included the same sequence of ZupT from C. difficile as in Figure S10A  discovered the presence of a nonsense mutation in SSA_1414 (W139*) in JFP227. SSA_1414 is 457 annotated as MutT, an 8-oxo-dGTP diphosphatase. We confirmed this mutation by Sanger 458 sequencing and found that it was unique to this strain, as it was not present in the Δ tmpA single 459 mutant (JFP226), which was made with the same PCR product as JFP227. We recreated 460 markerless versions of this mutation in the WT and Δ ssaACB backgrounds and found no 461 significant differences in serum growth for either version ( Figure S12). We next generated a 462 clean Δ ssaACB Δ tmpA strain and confirmed by Sanger sequencing that the SSA_1414 gene was 463 intact. We compared the growth of this new mutant (JFP377) to that of JFP227 ( Figure S13A-C) 464 under various conditions. There was no significant difference in growth between these strains. 465 We then measured the metal content of cells grown in BHI with 10 µM Mn 2+ using ICP-OES 466 and observed no significant difference in any metal examined ( Figure S13D). 467

468
With this study, it has now been confirmed that at least two different bacterial species, S. 469 sanguinis and B. burgdorferi, utilize a ZIP-family protein for manganese uptake. This firmly 470 establishes the ZIP family as an additional bacterial manganese importer family. Additionally, 471 we confirm that an NRAMP-family protein, MntH, contributes to manganese uptake in at least 472 one S. sanguinis strain in which it is naturally encoded. These findings confirm the existence of 473 secondary manganese transporters in S. sanguinis (Figure 12). Despite the fact that there was an 474 unintended mutation in the Δ ssaACB Δ tmpA mutant in a nearby gene, SSA_1414, we were able 475 to demonstrate that a clean mutant shares the same phenotype ( Figure S12) and that the Δ ssaACB 476 Δ tmpA mutant could be complemented by expression of the tmpA gene at a different 477 chromosomal site (Figure 3). 478 We showed that the Δ ssaACB mutant accumulates less iron and manganese, in agreement with 479 our previous results (Crump et al., 2014, Murgas et al., 2020) The effects of the Δ tmpA mutation in relation to iron are less easily understood than the effects 493 on manganese. We found that, as with manganese, the deletion of tmpA from the Δ ssaACB 494 mutant resulted in iron being less effective at restoring serum growth; indeed, while ~20-fold 495 more manganese was required to restore growth to the double mutant relative to its parent, no 496 amount of added iron (up to 1 mM) was sufficient to restore growth. This result could suggest 497 that TmpA is a secondary transporter of iron as well as manganese and that unlike manganese, 498 which we presume is transported into the double mutant by one or more tertiary transporters, 499 there is no tertiary transporter for iron under these conditions. Yet, unlike with manganese, the 500 Δ ssaACB Δ tmpA double mutant and its Δ ssaACB parent were indistinguishable with regard to 501 iron uptake in the presence or absence of supplemental iron (Figure 4), which is inconsistent with 502 TmpA-mediated iron transport. 503 We suggest that these seemingly inconsistent findings can be explained by hypothesizing that 504 there are certain core functions that require manganese, and no amount of iron can restore growth 505 if manganese levels are insufficient to meet these needs. Therefore, the failure of iron to 506 complement the double mutant is due to the loss of TmpA-mediated manganese transport rather 507 than TmpA-mediated iron transport. If we presume that increased levels of manganese are 508 needed for the core functions as levels of oxygen and, therefore, oxygen-dependent reactive 509 oxygen species such as superoxide and hydrogen peroxide increase (Anjem & Imlay, 2012), this 510 would also be consistent with the data in Figure 1 showing that as oxygen levels decrease, the 511 Δ ssaACB mutant is capable of ever more growth in serum. It should also be noted that 512 particularly in blood if not also in the oral cavity, free iron is present in vanishingly low levels 513 (Diaz-Ochoa et al., 2014); thus, the significance of iron transport by the SsaACB transporter is 514 not clear. 515 Despite the fact that most ZIP-family proteins transport zinc, there were no significant 516 differences between any Δ tmpA mutant and its respective parent in either zinc levels or zinc-517 deficient growth. Any difference in zinc levels between the Δ adcC and Δ adcC Δ tmpA mutants 518 with 10 µM added Zn 2+ was insignificant and given that this concentration is likely biologically 519 unattainable, the combined data suggest that zinc transport, if any, is minimal. 520 We have been unable to find a differential phenotype for the Δ tmpA mutant as compared to WT. 521 In our accompanying manuscript (Puccio et al., 2021b), we tested its growth in low-pH BHI and 522 saw no significant differences. We also assessed the growth in a biofilm assay in competition 523 with S. mutans observed no difference (data not shown). These results, along with those 524 described in this study, strongly suggest that its function is secondary to SsaACB. 525 The reason that most organisms encode multiple transporters for each metal is still contested, 526 although it highlights the importance of these transition metals. found that loss of Δ ssaACB in SK36 was detrimental to growth in low-pH media, which we 534 hypothesize is due to acid sensitivity of TmpA. In VMC66, lack of Δ ssaACB had no effect in the 535 same conditions, likely because MntH is still functional at the pH tested. Further studies will be 536 required to confirm the identity and selectivity of other metal transporters in S. sanguinis. 537 In VMC66, the only single manganese transport system mutant that showed significantly 538 decreased virulence was the Δ ssaACB mutant ( Figure 11A). The Δ ssaACB Δ tmpA mutant was 539 recovered at significantly lower levels than the Δ ssaACB mutant whereas the Δ ssaACB Δ mntH 540 mutant was not. However, when the two double mutants were compared to each other, they were 541 not significantly different ( Figure 11B). Thus, we cannot confidently conclude that the Δ tmpA 542 mutation had a greater effect on the virulence of the Δ ssaACB mutant than the Δ mntH mutation 543 did. 544 Expression of tmpA was not significantly affected by addition of metals or by depletion with 545 EDTA. EDTA is a non-specific metal chelator which was chosen to represent metal-deplete 546 conditions. While EDTA is not specific for any metal, it has an affinity for manganese (log β 1 of 547 14.1 and 24.8 for Mn 2+ and Mn 3+ , respectively, as compared to 16.7 for Zn 2+ and 8.7 for Mg 2+ ) 548 (Perrin & Dempsey, 1974). When added to WT or Δ ssaACB cultures growing in a fermentor, 549 ICP-OES analysis revealed that its primary effect on cellular metal levels was to reduce the 550 concentration of manganese (Puccio et al., 2020). Thus, if tmpA were regulated by manganese 551 depletion, we would have expected to see a change in expression after the addition of EDTA. 552 While our transcriptional analysis is not exhaustive, it is plausible that tmpA expression is 553 constitutive in S. sanguinis. In E. coli, the gene encoding ZupT is constitutively transcribed at 554 low levels (Grass et al., 2005). In our RNA-seq analysis of EDTA-treated cells, expression of 555 tmpA decreased slightly, yet significantly at T 50 but initial expression at T -20 was low, as were 556 other genes in the operon . This result may have been an artifact of data 557 normalization, because in our study examining the transcriptome after pH reduction, expression 558 of tmpA remained low and constant in both WT and Δ ssaACB strains, despite a decrease in 559 cellular manganese levels (Puccio et al., 2021). 560 In a recent study, Zhang et al. (2020) confirmed that in the human hZIP4, the M1 site is essential 561 for metal transport and the M2 site facilitates optimal transport activity. Here we report that the 562 modification of N173 to A or D in the predicted M2 site of TmpA resulted in the reduced 563 function of the protein as a manganese transporter (Figure 8). This indicates TmpA cannot 564 function efficiently without an asparagine in this position, despite the similar size and features of 565 the aspartic acid. These results suggest that the charge of the side chain at this position in TmpA 566 likely contributes to either transport function, metal selectivity, or both. Future studies using cell-567 free metal uptake assays and structural analysis will be required to confirm our results. 568 The discovery that streptococci have different lineages of ZIP-family proteins suggests that these 569 secondary transporters may be valuable to some species. However, this protein family is not 570 essential in all streptococci as some species, such as the human pathogen S. pyogenes and the 571 oral commensals S. oralis and S. mitis, have been evolutionarily successful without a ZIP-family 572 protein.
In some instances, MntH is encoded in the genomes of these species but S. pyogenes has 573 only two sequenced strains that encode a MntH homolog, suggesting that the other strains either 574 rely entirely on the primary ABC manganese transporter or encode an unidentified secondary 575 transporter. At the other extreme, some streptococci encode all three transporters, similar to S. 576 sanguinis VMC66 examined here. S. mutans is an example; however, the contribution of ZupT Sm 577 to manganese transport has yet to be examined. A recent study found that it did not contribute 578 significantly to zinc transport (Ganguly et al., 2021). Another study from the same group 579 utilizing S. mutans manganese-transport mutants deleted for sloC, the ssaB homolog, and mntH 580 found that the inactivation of MntH alone did not produce any obvious phenotype, whereas the 581 Δ sloC and Δ sloC Δ mntH mutants were severely deficient under manganese-restricted conditions 582 (Kajfasz et al., 2020). Further studies in both S. mutans and S. sanguinis would improve the 583 understanding of the relative contribution of each of these proteins to growth, metal transport, 584 and virulence and possibly reveal more about the origin of the distinct lineages. 585 In conclusion, we discovered that this ZIP-family protein, TmpA, contributes to manganese 586 uptake and virulence in several strains of S. sanguinis, which is evident only after the deletion of 587 the primary transporter. These results lay the foundation for future studies of manganese-588 transporter inhibitors for the prevention of IE, as both transporters could be targeted 589 simultaneously to prevent the generation of spontaneous mutants that could subvert single-target 590 treatment. Indeed, a study targeting BmtA to prevent B. burgdorferi virulence showed promising 591 results (Wagh et al., 2015), although this study was completed before the crystallization of any 592 ZIP-family protein. Gene knockout mutant constructs were either generated previously  or by gene 622 splicing by overlap extension (SOEing) PCR (Ho et al., 1989) where the gene(s) of interest were 623 replaced with an antibiotic resistance gene or cassette. Transformations were performed using 624 the protocol described previously (Paik et al., 2005 Overnight cultures of SK36 WT and Δ ssaACB (JFP169) strains were grown as described above. 720 They were then diluted 10-fold into BHI and incubated aerobically (~21% O 2 ). Once cells 721 reached mid-log phase (OD 600 ~0.6), 6 mL of culture was separated into tubes for each condition 722 and 100 µM Puratronic TM metal (MnSO 4 , ZnSO 4 , or FeSO 4 ) or EDTA (Invitrogen) was added. A 723 culture tube with no additives was included as the control. Tubes were incubated aerobically 724 without a jar at 37°C for 15 min. To collect cells, the tubes were swirled in a dry ice/ethanol bath 725 for 30 s prior to centrifugation for 10 min at 3,740 x g at 4°C. The supernatant was discarded and 726 the samples stored at -80°C. In some experiments, overnight cultures of SK36 and JFP169 were 727 grown in an anaerobic chamber (Coy Laboratory Products). Cells were then diluted 10-fold into 728 BHI pre-incubated anaerobically at 37°C. At mid-log phase, cultures were separated and 6 mL of 729 cells were collected immediately and 6 mL of cells were incubated aerobically (~21% O 2 ) for 15 730 min. Cells were collected as above. RNA isolation and on-column DNase treatment were 731 completed using the RNeasy Mini Kit and RNase-Free DNase Kit, respectively (Qiagen). RNA 732 was eluted in 50 µL RNase-Free water (Qiagen). A second DNase treatment was then performed 733 on the samples (Invitrogen). Total RNA was quantified and purity was assessed using a 734 Nanodrop 2000 Spectrophotometer (ThermoScientific). Libraries of cDNA were created using 735 SensiFAST cDNA Synthesis Kit (Bioline). Control reactions without reverse transcriptase were 736 conducted to confirm the absence of contaminating DNA in all samples. Quantitative real time Positions within a cellular membrane were predicted using OPM (https://opm.phar.umich.edu/) 780 and visualized in JMol 3.0 using FirstGlance (http://jmol.sourceforge.net/). 781

Evolutionary analysis 782
Sequences of S. sanguinis SK36 TmpA and SsaB and S. mutans UA159 TmpA were used as 783 queries for BLASTP searches of the Non-redundant protein or Refseq Select proteins database 784 and the top hits were selected. For each species selected, only one sequence was used for the 785 phylogenetic analysis. The evolutionary history was inferred using the Neighbor-Joining method 786 (Saitou & Nei, 1987) with a gap opening penalty of 3.0 and a gap extension penalty of 1.8 for the 787 multiple alignment stage (Hall, 2008). The evolutionary distances were computed using the 788 Poisson correction method (Zuckerkandl & Pauling, 1965)  and NRAMP-Family Protein. Red arrows indicate the direction of manganese movement. Figure  820 made with Biorender. Cultures were grown in (A) rabbit serum or (B) BHI at the given O 2 concentration for 24 h. 1093 Means and standard deviations of at least three independent experiments are displayed. 1094 Significant differences between each