Protein Interactomes Identify Distinct Pathways for Streptococcus mutans YidC1 and YidC2 Membrane Protein Insertases

Virulence properties of cariogenic Streptococcus mutans depend on integral membrane proteins. Bacterial protein trafficking involves the co-translational signal recognition particle (SRP) pathway components Ffh and FtsY, the SecY translocon, and membrane-localized YidC chaperone/insertases. Unlike Escherichia coli, S. mutans survives loss of the SRP pathway. In addition, S. mutans has two yidC paralogs. The ΔyidC2 phenotype largely parallels that of Δffh and ΔftsY while the ΔyidC1 phenotype is less severe. This study defined YidC1 and YidC2 interactomes to identify their respective functions alone and in concert with the SRP, ribosome, and/or Sec translocon. A chemical cross-linking approach was employed, whereby whole cell lysates were treated with formaldehyde followed by Western blotting using anti-Ffh, FtsY, YidC1 or YidC2 antibodies and mass spectrometry (MS) analysis of gel-shifted bands. Cross-linked lysates from WT and ΔyidC2 strains were also reacted with anti-YidC2 antibodies coupled to magnetic Dynabeads™, with co-captured proteins identified by MS. Additionally, C-terminal tails of YidC1 and YidC2 were engineered as glutathione-S-transferase fusion proteins and subjected to 2D Difference Gel Electrophoresis and MS analysis after being reacted with non-cross-linked lysates. Results indicate that YidC2 works in concert with the SRP-pathway, while YidC1 works in concert with the SecY translocon independently of the SRP. In addition, YidC1 and/or YidC2 can act alone in the insertion of a limited number of small integral membrane proteins. The YidC2-SRP and YidC1/SecY pathways appear to function as part of an integrated machinery that couples translation and transport with cell division, as well as transcription and DNA replication. Importance Streptococcus mutans is a prevalent oral pathogen and causative agent of tooth decay. Many proteins that enable this bacterium to thrive in its environmental niche, and cause disease, are embedded in its cytoplasmic membrane. The machinery that transports proteins into bacterial membranes differs between Gram-negative and Gram-positive organisms. One important difference is the presence of multiple YidC paralogs in Gram-positive bacteria. Characterization of a protein’s interactome can help define its physiological role. Herein, we characterized the interactomes of S. mutans YidC1 and YidC2. Results indicate that YidC1 and YidC2 have individualized functions in separate membrane insertion pathways, and suggest putative substrates of the respective pathways. Furthermore, S. mutans membrane transport proteins appear as part of a larger network of proteins involved in replication, transcription, translation, and cell division/cell shape. This information contributes to our understanding of protein transport in Gram-positive bacteria in general, and informs our understanding of S. mutans pathogenesis.


Introduction
secreted SecA substrates, although SecA itself would be in proximity of SRP components on the 197 ribosome. Also of interest, the middle region reactive with both anti-YidC1 and anti-YidC2 198 antibodies, did not include SecY, YajC, or SRP components. E. coli YidC is known to function 199 in a YidC only pathway in the insertion of small membrane proteins [13,22,45]. Interestingly, 4 200 out of the 5 membrane proteins with only 1 or 2 TM domains were present in the middle region 201 (Table 1). This suggests that YidC1 and/or YidC2 can work independently of SecY, and the 202 SRP pathway, in the insertion of a limited number of small membrane proteins. 203 In an attempt to identify respective substrates of YidC1-SecY, YidC1/2, or SRP-YidC2 204 mediated pathways, those proteins predicted to have one or more transmembrane were 205 categorized according to their presence in upper, middle, or lower regions of the gel (Table 1). 206 The identification of a higher proportion of membrane proteins in the lower region suggests that 207 YidC1-SecY/YajC is a widely used pathway for membrane protein insertion. That is, YidC1 208 likely represents the "housekeeping" paralog in S. mutans. A closer examination of membrane 209 proteins in the lower region revealed known or putative metal transporters including 210 SMU_770C, SMU_998, and a putative zinc ABC transporter ATP-binding protein (SMU_1994) 211 suggesting that these particular metal transporters utilize the YidC1-SecY/YajC pathway for 212 insertion. Also, proteins encoded in an operon of unknown function that includes SMU_832, 213 SMU_833, and SMU_834 were identified in the lower region. Proteins in the upper region, 214 suggestive of insertion by a coordinated SRP-YidC2 pathway, included multiple sugar 215 transporters and several ABC transporters including the competence-associated protein ComA. 216 S. mutans deletion mutants lacking ffh or yidC2 exhibit impaired genetic competence, but this 217 property is less impacted by elimination of yidC1 [46]. We did not observe a preference for 218 single TM compared to multi-pass membrane proteins for the YidC1-SecY/YajC or the SRP-YidC2 pathways. In contrast, as stated above, membrane proteins from the middle region that 220 may represent substrates of a YidC1 and/or YidC2 only pathway were mostly single or double-221 pass membrane proteins except for hemolysin (SMU_1693), which contains 4 predicted TM 222 domains. 223 While in vivo whole cell cross-linking identified a relatively low number of potential 224 membrane-localized substrates of the putative SRP-YidC2 mediated protein translocation 225 pathway, a high proportion of proteins (30/65) present in the upper region of the gel are involved 226 in DNA replication/repair, transcription, translation, and cell division/cell shape (Table 1). 227 These results support the idea that the SRP-YidC2 co-translational translocation pathway 228 operates in the context of a larger consortium of proteins that make up an integrated higher order 229 machinery, which couples replication, transcription, and cell division with membrane insertion of 230 a subset of membrane proteins. In contrast, approximately 25% of proteins present in the lower 231 gel slice representative of the putative SecY-YajC /YidC1 pathway are membrane proteins. 232 Additionally, multiple proteins identified in this region are associated with replication, 233 transcription, or translocation. The streptococcal homolog of the ribosome-associated chaperone 234 trigger factor, RopA, was identified in both the upper and lower regions that likely represent the 235 SRP-YidC2 and SecY-YidC1-mediated protein translocation pathways, respectively. Trigger 236 factor has been reported to bind to the same ribosomal protein at the peptide exit site as the SRP 237 pathway [47,48]; therefore, the finding of RopA and Ffh in the same gel slice was not 238 unexpected. 239 Because a whole cell cross-linking approach is limited by the accessibility of exposed 240 functional groups in the target proteins to formaldehyde [49], the actual integral membrane 241 substrates of the insertion machinery were likely underrepresented in our dataset due largely to 242 their being buried within the membrane and inaccessible to the cross-linking reagent. 243 Surprisingly YidC1, YidC2, Ffh, and FtsY themselves were also not identified in either cross-244 linked or non-cross-linked samples, although they were clearly present as evidenced by their 245 detection by Western blot and migration at the correct molecular weight. It is possible that they 246 were not amenable to or present in sufficient quantity for detection by mass-spectrometry. 247 Western blot with high quality antibodies can be more sensitive than standard bottom-up MS in 248 the detection of certain proteins of low abundance [50]. To overcome this potential limitation, 249 we also employed immunocapture experiments in an attempt to improve sensitivity. 250 Dynabead™ immunocapture of protein complexes from S. mutans lysates using anti-YidC2 251 antibodies. To identify potential binding partners of YidC2 and to characterize this insertase's 252 interactome, an immunocapture approach was undertaken in which anti-YidC2 antibodies were 253 covalently coupled to magnetic Dynabeads™. The polyclonal rabbit antibodies used were made 254 against synthetic peptides corresponding to the YidC2 C-terminal tail and cytoplasmic loop 255 between TM2 and TM3. Whole cell lysates from untreated and formaldehyde-treated cells of S. 256 mutans strain NG8, and its corresponding yidC2 mutant, were reacted with the antibody-257 coupled beads and bound proteins were eluted with glycine-HCl, pH 2.0. Aliquots of each 258 sample were analyzed by Western blot ( Fig. 2A). As expected, a 27 kDa YidC2 band was 259 identified in the wild-type (WT), but not yidC2 strain (Fig. 2B). An additional band reactive 260 with anti-YidC2 antibodies was also observed in the cross-linked sample from the WT, but not 261 samples from the mutant strain, or the non-cross-linked control sample from the WT strain. 262 Other higher molecular weight bands were observed in a replicate negative control blot probed 263 only with goat-anti-rabbit heavy chain specific secondary antibodies. These represent anti-264 YidC2 antibodies present in the eluate that leached from the coupled Dynabeads™ (Fig. 2C).
Gel slices corresponding to the ~45 kDa region of interest identified by Western blot were cut 266 from SDS-PAGE gels of all four samples and analyzed by MS. A total of 269 proteins were 267 identified in the WT cross-linked sample, 229 in the WT non-cross-linked sample, 246 proteins 268 in the yidC2 cross-linked sample, and 284 proteins in the yidC2 non-cross-linked sample.

269
Sixty-eight proteins were present in the WT cross-linked sample and 31 in the non-cross-linked 270 sample that were absent from the corresponding yidC2 samples (summarized in Table S2). Six 271 of those were shared between cross-linked and non-cross-linked samples. The presence of such 272 shared proteins may represent direct binding partners of YidC2 that do not need to be cross-273 linked to be co-captured with it. A graphical representation of the types of proteins co-captured 274 with YidC2 is shown (Fig. 2D). 275 Twenty-four of the 99 total immunocaptured proteins identified in WT but not yidC2 276 samples are predicted to have one or more TM domains, with the rest having been identified as 277 membrane-associated in our previous membrane proteomic analyses [10] (Table S2). Potential 278 integral membrane substrates of YidC2 identified by this immunocapture experiment include two 279 subunits of the PTS mannose transporter, metalloprotease (RseP), histidine kinases, enzymes, 280 and cell wall/cell division related proteins (Table 2). In agreement with our prior gel shift 281 experiment and analysis of the upper gel slice, we identified SecA as well as other proteins 282 involved in DNA replication/repair, transcription, translation, and cell division/cell shape in 283 associated with YidC2 (Fig. 2D, Panel D). Again this suggests that translocation is part of a 284 coordinated machinery that incorporates additional processes beyond protein translation. In 285 contrast to the gel shift assay, in which both YajC and SecY were detected in the lower gel slice 286 reactive with anti-YidC1 antibodies, in the current immunocapture experiment YajC, but not 287 SecY, was identified as part of YidC2 interactome. Certain differences in the apparent YidC2 interactome identified following Western blot gel shift, as opposed to Dynabead TM 289 immunocapture, may relate to those domains of YidC2 available for binding to substrates or 290 other proteins during the two experimental approaches. Antibodies against cytoplasmic loop 1 291 and the YidC2 tail domain were utilized for Dynabead TM capture of YidC2 and associated 292 proteins. Thus binding of proteins that interact specifically with either of these regions may have 293 prevented efficient capture of YidC2 by the antibody-coupled magnetic beads. That is, the 294 immunocapture dataset was likely biased against proteins that react with cytoplasmic loop 1 or 295 the YidC2 C-terminal tail. In E. coli, cytoplasmic loop 1 of YidC has been reported to interact 296 with SecY in an in vivo photo-crosslinking assay [17]. Thus, occupancy of the corresponding 297 loop in YidC2 by SecY could potentially have blocked reactivity with the anti-YidC2 loop 298 antibody and explain why YajC, but not SecY, was identified in the immunocapture assay. 299 Unlike the Western blot gel shift results, we did not identify Ffh or FtsY in association with 300 YidC2 in the immunocapture experiment. If the cooperative activity of YidC2 with the SRP 301 pathway depends on an interaction mediated by its C-terminal tail, that could preclude its 302 efficient capture by anti-tail-specific antibodies. Previous domain swapping experiments support 303 this conjecture in that stress tolerance was complemented in a yidC2 background with chimeric 304 YidC1 whose C-terminal tail was replaced with that of YidC2 [8]. Collectively, our anti-YidC2 305 immunocapture assay identified not only the translocation machinery components SecA and 306 YajC, but also ribosomal proteins, chaperones and proteases, enzymes involved in DNA 307 replication and repair, and proteins responsible for cell wall generation and cell division. 308 Because of issues with low coupling efficiency of the anti-YidC2 antibodies to Dynabeads TM and 309 the problem with antibody leaching, this approach was not attempted with anti-YidC1 antibodies.

Difference gel electrophoresis of S. mutans proteins captured by YidC1 or YidC2 C-311
terminal tails. As an alternative to immobilization of anti-YidC antibodies to Dynabeads TM , we 312 also utilized a Glutathione-S-transferase (GST)-tagged based pull-down approach. While it is 313 difficult to express S. mutans yidC1 and yidC2 in E. coli to sufficient levels for large scale 314 protein purification, both the YidC1 and YidC2 C-terminal tails are soluble, and easily tagged 315 and purified. We constructed fusion proteins of the YidC1 and YidC2 C-terminal domains with 316 GST and affinity purified the recombinant polypeptides on Glutathione Sepharose TM (Fig. S1). 317 Because domain swapping experiments have demonstrated that the positively-charged tails of S. 318 mutans YidC1 and YidC2 contribute to certain functional attributes of each paralog [8], we 319 expected a subset of YidC1 and YidC2 binding partners to interact with these domains. After 320 purification, the GST-tagged YidC1/2-tail fusion proteins were reacted with S. mutans whole cell 321 lysates (non-cross-linked) and captured on immobilized glutathione using GST as a negative 322 control. Following elution with reduced glutathione the three samples were individually labeled 323 with a different CyDye fluorescent dye and subjected to 2D-difference gel electrophoresis 324 (DIGE) (Fig. 3). One hundred and twenty-one spots were identified as being captured by  YidC1CT and/or GST-YidC2CT, but not by GST (Fig. S2). A complete list of all proteins 326 identified in each of the gel spots is shown in Table S3. A summary of the proteins pulled down 327 with GST-Yid1CT (green spots), GST-YidC2CT (red spots), or both (yellow spots) is shown in 328 Table S4. Seventy-four proteins were co-captured with GST-YidC1CT, and 37 with GST-329 YidC2CT (Table S3). Of those, 42 were uniquely co-captured with GST-YidC1CT, while only 5 330 were uniquely co-captured with GST-YidC2CT (Table S4). 331 The types of proteins co-captured with GST-YidC1CT compared to GST-YidC2CT are 332 summarized in Table 3. Proteins with 1 or more transmembrane domains were considered as 333 putative substrates. Eleven different integral membrane proteins were found as part of the 334 YidC1-tail interactome, including 5 that were also pulled down with GST-YidC2CT. Most of 335 these were transporters with the exception of the cell division protein FtsH, and a histidine kinase 336 (SMU_486). All non-integral membrane proteins identified by DIGE (~85%) had previously 337 been identified as membrane-associated during proteomic analysis of S. mutans protoplast-338 derived membrane preparations [10]. The predominance of non-integral membrane proteins in 339 the DIGE dataset suggest that the YidC1 and YidC2 C-terminal tails do not play a prominent role 340 in recognizing and binding substrates. Twenty of the 24 membrane-associated proteins from the 341 YidC2-tail interactome were also co-captured with GST-YidC1CT. Interestingly, the SRP 342 component protein Ffh was found in association with the YidC2-tail, but not with the YidC1-tail. 343 This supports data from the in vivo cross-linking experiments that suggested a cooperative SRP-344 YidC2 pathway, and explains why appending the YidC2 tail onto YidC1 enables the chimeric 345 protein to ameliorate the yidC2 phenotype. Presumably this manipulation allows the YidC1 346 insertase to interact with Ffh and function in concert with the SRP pathway machinery. None of 347 the components of the SecYEG translocon, nor YajC, were identified in association with either 348 of the C-terminal tails, thus these domains likely do not contribute to YidC1 or YidC2 349 interactions with the translocon itself. 350 As described above in the gel shift and YidC2 immunocapture experiments, numerous 351 ribosomal proteins, as well as other components of the translation machinery, were captured in 352 association with YidC1 and/or YidC2. Such proteins were found irrespective of whether Ffh and 353 FtsY were also present, suggesting that either S. mutans YidC paralog can act to support co-354 translational protein translocation in the absence of the SRP pathway. Indeed, YidC2 was 355 previously demonstrated to complement Oxa1 deficiency in yeast mitochondria that lack an SRP pathway [27]. While YidC1 was present in yeast cell extracts, this paralog was not properly 357 imported into the mitochondria and therefore could not be assessed in complementation 358 experiments [27]. When overexpressed in E.coli, both YidC1 and YidC2 of S. mutans were 359 found to interact with translating and non-translating ribosomes by a tail-dependent mechanism 360 [51]. In the current study, the large ribosomal subunit protein, L2, was the most abundant 361 ribosomal protein pulled down by both the GST-YidC1CT and GST-YidC2CT fusion 362 polypeptides. In E. coli, L2 not only acts as a structural component of the ribosome, it is also 363 processed to a truncated derivative (tL2)  was cloned and the recombinant his-tagged proteins were tested by ELISA to determine whether 369 either form interacts directly with the C-terminal tails of YidC1 or YidC2. Neither L2 nor rL2 370 demonstrated significant binding to GST-YidC1CT or to GST-YidC2CT (Fig. S3). Likewise, 371 SecA, which had been observed in conjunction with YidC2 in both Western blot gel-shift and 372 immunocapture experiments, did not react directly with GST-YidC2CT (or GST-YidC1CT) (Fig. 373 S3). This suggests that the association of SecA with YidC2 is indirect, or nor mediated by the 374 YidC2 tail. 375 Similar to the previous experiments, GST-YidC1CT and GST-YidC2CT also captured a 376 variety of proteins including chaperones and those involved in replication, transcription, 377 translation, and cell division/cell shape again suggesting that all these processes are temporally 378 and spatially connected. These data are consistent with the identification of coupled 379 transcription/translation in other bacteria, which may also integrate aspects of DNA replication 380 [55][56][57][58][59]. Both YidC1 and YidC2 contribute to proper cell wall biosynthesis and cell morphology 381 in S. mutans [9], thus capture of proteins in this category is consistent with previously described 382 mutant phenotypes. 383 Determination of YidC1 and YidC2 interactomes and functional annotation. When proteins 384 from all experiments were evaluated in composite, 88 were identified as being associated with 385 both YidC1 and YidC2, while 123 or 131 were uniquely associated with YidC1 or YidC2, 386 respectively (Fig. 4A). When possible, proteins were assigned to functional categories by 387 DAVID analysis (Fig. 4B). The most prevalent functional category in both interactomes was 388 transferase. Functional annotation also shows that the YidC2, compared to YidC1, interactome 389 was enriched in a number of functional categories including ATP-binding proteins, 390 metalloproteins, carbon metabolism, oxidoreductases, cell division, GTPase activity, and 391 branched chain amino acid pathways. This may explain why the phenotypic consequence of 392 elimination of YidC2 is far more pronounced than elimination of YidC1 [7,8,46]. In contrast, 393 the only instances in which the YidC1 interactome equaled or exceeded that of YidC2 were in 394 the transferase, and purine and pyrimidine metabolism categories. Of note, however, a greater 395 number of proteins in the YidC1 interactome are either not annotated or have putative 396 individualized functions that cannot be assigned to a broad category. 397 We also carried out a protein-protein interaction (PPI) network analysis using the 398 STRING (Search Tool for the Retrieval of Interacting Genes/Proteins). YidC1 and YidC2, as 399 well as all proteins experimentally identified as associating with either or both of them, were 400 included in the uploaded datasets. The individual YidC1 and YidC2 STRING interactomes are 401 shown in Fig. 5A and 5B, and the common interactome in Fig 5C. The majority of the proteins we identified in the current study were included within the PPI networks predicted by STRING, 403 thus giving us high confidence in the accuracy of the experimentally determined protein 404 interactomes. Consistent with co-translational protein translocation pathways, the most intense 405 nodes identified in all three PPI network predictions were largely comprised of ribosomal 406 proteins and other components of the translation machinery. L2, which we determined by 407 ELISA not to interact with the YidC C-terminal tails (Fig. S3), was not predicted by STRING 408 analysis to interact with either YidC1 or YidC2. S1 however, is a predicted STRING interaction 409 partner of L2, as well as of YidC1 and YidC2. S1 is therefore a likely bridging molecule since it 410 was detected experimentally whenever L2 was found in association with either YidC1 or YidC2. 411 3 M ammonium sulfate. Following incubation, the tube was placed next to a magnet and the 489 supernatant removed. Unbound antibodies were removed from the beads by washing first with 1 490 ml PBS containing 0.5% TritonX-100, and secondly with freshly made 0.5 N NH4OH, 0.5 mM 491 EDTA, until A280 of the wash supernatant was zero. Ab-coated beads were washed three times 492 with 1 ml PBS, resuspended in 100 l PBS, and reacted with ~700 ul formaldehyde cross-linked 493 whole cell lysate samples (~ 8 mg/ml) derived from S. mutans strain NG8 (wild type) or PC398 (yidC2), or with control samples prepared without formaldehyde, for three hours at 4 C with 495 gentle end over end rotation. Next, beads were separated with a magnet and washed six times 496 with 1 ml PBS. Ab-captured proteins were eluted with 0.5 ml freshly made 0.5 N NH4OH, 0.5 497 mM EDTA, and vortexing in an Eppendorf tube adapter (Vortex Mixer, Fisher Scientific) set at 498 medium speed for 20 min at RT. Beads were removed with a magnet and the eluate was snap-499 frozen in liquid nitrogen and dried overnight at RT in a SpeedVac vacuum concentrator (Savant, 500

Concluding Remarks and Apparent
Famingdale, NY). Twenty microliters of SDS sample buffer (62.5 mM Tris pH 6.8, 10 % 501 Glycerol, 0.2 % SDS, 0.02% Bromphenol Blue) were added to each dried sample and incubated 502 for 10 min at 65C. The samples were electrophoresed on 4-20% gradient SDS-polyacrylamide 503 gels (Bio-Rad, Hercules, CA) and analyzed by Western blot using anti-YidC2 C-terminal-504 specific antibodies as described above. Controls included non-cross-linked samples prepared 505 without formaldehyde, and a Western blot developed with HRP-conjugated goat anti-rabbit 506 secondary antibody only. 507 Preparation of gel slices for protein identification by mass spectrometry. SDS 508 polyacrylamide gels were rinsed in Optima LC-MS grade water (Fisher Scientific) three times, 509 fixed for 15 min with 50% methanol and 7% acetic acid (Fisher Scientific), and stained with 510 GelCode, Blue Stain Reagent (Thermo Scientific) according to the manufacturer's instructions. 511 Gel slices corresponding to gel-shifted regions identified by Western blot with anti-YidC1, 512 YidC2, Ffh or FtsY-specific antibodies in the formaldehyde cross-linked UA159 whole cell 513 lysate, but absent from the non-cross-linked control sample, were excised for in situ proteolysis. 514 Similarly, a band detected by Western blot with anti-YidC2 antibodies in the DynaBead™ eluate 515 of the NG8 formaldehyde cross-linked sample, but not the yidC2 mutant strain or non-cross-516 linked control samples, was excised for proteolysis from the same location of SDS-517 polyacrylamide gels of all four samples. Gel slices were washed twice in nanopure water for 5 518 minutes, then destained with 1:1 v/v methanol: 50 mM ammonium bicarbonate for ten minutes 519 with two changes. Gel slices were dehydrated with 1:1 v/v acetonitrile: 50 mM ammonium 520 bicarobonate, then rehydrated and incubated with dithiothreitol (DTT) solution (25 mM in 100 521 mM ammonium bicarbonate) for 30 minutes prior to the addition of 55 mM Iodoacetamide in 522 100 mM ammonium bicarbonate solution. Gel slices were incubated for an additional 30 min in 523 the dark then washed with two cycles of water and dehydrated with 1:1 v/v acetonitrile: 50 mM 524 ammonium bicarbonate. Protease was driven into the gel pieces by rehydrating them in 12 ng/ml 525 trypsin in 0.01% ProteaseMAX Surfactant (Promega) for 5 minutes. Gel pieces were next 526 overlaid with 40 µL of 0.01% ProteaseMAX surfactant: 50 mM ammonium bicarbonate and 527 gently mixed on an orbital shaker for 1 hour. The digestion was stopped by addition of 0.5% 528 trifluoroacetic acid. MS analysis was performed immediately to ensure high quality tryptic 529 peptides with minimal non-specific cleavage. 530 Mass spectrometry analysis. Nano-liquid chromatography tandem mass spectrometry (Nano-531 LC/MS/MS) was performed on a Thermo Scientific Q Exactive HF Orbitrap mass spectrometer 532 equipped with an EASY Spray nanospray source (Thermo Scientific) operated in positive ion 533 mode, or on a Quadrupole-Tof (Q-TOF) instrument. The LC system was an UltiMate™ 3000 534 RSLCnano system from Thermo Scientific. The mobile phase A was water containing 0.1% 535 formic acid acetic acid and the mobile phase B was acetonitrile with 0.1% formic acid. Five 536 microliters of each sample was first injected on to a Thermo Fisher Scientific Acclaim Trap 537 Cartridge (C18 column, 75 um ID, 2 cm length, 3 m 100 Å pore size) and washed with mobile 538 phase A to desalt and concentrate the peptides. The injector port was switched to inject and the 539 peptides were eluted off of the trap onto the column. An EASY Spray PepMAP column from m 100 Å pore size). The column temperature was maintained 35 C as peptides were eluted 542 directly off the column into the LTQ system using a gradient of 2-80%B over 45 minutes, with a 543 flow rate of 300 nL/min. The total run time was 60 minutes. were accepted if they could be established at greater than 99.0% probability and contained at 565 least 1 identified peptide. Protein probabilities were assigned by the Protein Prophet algorithm 566 [65] . 567 Two-dimensional difference gel electrophoresis (DIGE) analysis of S. mutans proteins 568 captured with GST-YidC1 compared to GST-YidC2 C-terminal tail fusion proteins. The 569 C-terminal fragment (bp682-816) of yidC1 was amplified by PCR using primers NL5F 570 (ggaacggatcccaggtcttccagattctgttg) and NL5R (ccgtagtcgacttattttctctttttatgtgctttc). The C-571 terminal fragment (bp742-933) of yidC2 was amplified by PCR using primers NL6F 572 (ggaacggatccacaaaccatatcattaaaccaaaat) and NL6Rb (ccgtagtcgacttattgcttatggtgacgctgt). S. 573 mutans UA159 genomic DNA was used as the template. PCR products were digested with 574 BamHI and SalI and ligated to corresponding restriction enzyme sites in the pGEX-4T-2 vector. 575 The vector only encoding GST was transformed into BL21 DE3 (ThermoFisher Scientific). 576 Plasmids encoding GST-YidC1CT or GST-YidC2CT were transformed into BL21 Star™ 577 (ThermoFisher). GST-YidC1CT expression was induced with 0.5 mM isopropyl-ß-D-578 thiogalactopyranoside (IPTG) for 6 hours at 30 C. Expression of GST and GST-YidC2CT was 579 induced with 1 mM IPTG for 4 hours at 37 C. Cells were harvested by centrifugation at 11,325 580 x g for 15 min, and resuspended in 25 ml PBS. Cell suspensions were supplemented with 1 mM 581 phenylmethylsulfonyl fluoride (PMSF) (Acros Organics) and protease inhibitor cocktail (1 mini 582 tablet/25 ml) (Roche Diagnostics GmbH). Cell lysis was performed using an Avestin 583 EmulsiFlex-C5 high-pressure homogenizer (Avestin Inc., Ottawa, Ontario, Canada) at a pressure 584 of 15,000-20,000 p.s.i. for three cycles. Cell debris was removed by centrifugation at 11,000 x g 585 for 30 min and the supernatants filtered through a 0.22 µm syringe filter (Merck Millipore).   30S ribosomal protein S3 (SMU_2021)* 30S ribosomal protein S10 (SMU_2026c) 30S ribosomal protein S13 (SMU_2003) 30S ribosomal protein S11 (SMU_2002) 30S ribosomal protein S18 (SMU_1858) Translation elongation factor EF-Tu (SMU_714) § Translation elongation factor G (SMU_359) Putative translation elongation factor TS (SMU_2031) Translation initiation factor 2 (SMU_421) Putative translation initiation factor IF3 (SMU_697) Putative alanyl-tRNA synthetase (alanine--tRNA ligase) (SMU_650)