The Acinetobacter baumannii Mla system and glycerophospholipid transport to the outer membrane

The outer membrane (OM) of Gram-negative bacteria serves as a selective permeability barrier that allows entry of essential nutrients while excluding toxic compounds, including antibiotics. The OM is asymmetric and contains an outer leaflet of lipopolysaccharides (LPS) or lipooligosaccharides (LOS) and an inner leaflet of glycerophospholipids (GPL). We screened Acinetobacter baumannii transposon mutants and identified a number of mutants with OM defects, including an ABC transporter system homologous to the Mla system in E. coli. We further show that this opportunistic, antibiotic-resistant pathogen uses this multicomponent protein complex and ATP hydrolysis at the inner membrane to promote GPL export to the OM. The broad conservation of the Mla system in Gram-negative bacteria suggests the system may play a conserved role in OM biogenesis. The importance of the Mla system to Acinetobacter baumannii OM integrity and antibiotic sensitivity suggests that its components may serve as new antimicrobial therapeutic targets.


INTRODUCTION 27
Gram-negative bacteria are enveloped by two lipid bilayers, separated by an aqueous 28 periplasmic space containing a peptidoglycan cell wall. The inner membrane (IM) is a symmetric 29 bilayer of glycerophospholipids (GPL), of which zwitterionic phosphatidylethanalomine (PE), 30 acidic phosphatidylglycerol (PG), and cardiolipin (CL) are among the most widely distributed in 31 We next used cryo-electron microscopy to characterize the architecture of the A. 155 baumannii MlaBDEF complex (abMlaBDEF). This complex is uniformly dispersed in vitreous 156 ice (Fig. S3A), and 2D classification demonstrated the presence of a range of views suitable for 157 structure determination (Fig. S3B). Following 2D-and 3D-classification, we obtained a final 158 dataset of ~ 14,000 particles with which we obtained a structure to a resolution of 8.7 Å ( independently, are identical, suggesting that they correspond to the correct structure for the 167 complex. However, the limited resolution of the ecMlaBDEF complex structure did not allow 168 modeling of its individual subunits, in contrast to the abMlaBDEF structure reported here. 169 We note that a clear six-fold symmetry is present for the region of the map attributed to 170 MlaD (Fig. 2B), despite the fact that we only imposed a 2-fold symmetry averaging. This agrees 171 with the proposed hexameric state of its E. coli homologue (ecMlaD) (19). We next modeled 172 abMlaD, using an evolution restraints-derived structural model of ecMlaD (21) as a template, 173 and used our previously-published EM-guided symmetry modeling procedure (22) to model its 174 hexameric state. The obtained abMlaD hexameric model is at a low-energy minimum (Fig. S4B) 175 and fits the EM map density well (Fig 2B and S5B Fig). A crystal structure of the periplasmic 176 domain of ecMlaD published recently (20) formed a crystallographic hexamer, suggesting that 177 this corresponds to the native hexomeric arrangement for this domain. Our abMlaD hexameric 178 model is very similar to the crystallographic ecMlaD structure (Fig. S4C), supporting the 179 proposed domain arrangement in the MlaBDEF complex. We note, however, that one region of 180 density in the EM map is not accounted for by our MlaD hexamer model (Fig. 2B). The 181 localization of this extra density suggests that it corresponds to a ~ 45 amino-acid insert present 182 between strands 4 and 5 of the abMlaD β-sheet (Fig. S5A). The role of this insert, uniquely 183 found in the A. baumannii orthologue, is not known. 184 We next modeled the structures of MlaB and MlaF and fitted their respective coordinates 185 in the corresponding region of the EM map ( Fig. 2C and Fig. S4A). For both proteins, most 186 helices are well resolved, which allowed us to place the models unambiguously. We then 187 compared the conformation of the ATPase MlaF to that of the maltose transporter ATPase MalK, 188 which has been trapped in several conformations of the transporter; i.e. the inward-facing state, 189 the pre-translocation state, and the outward-facing state (23,24). Interestingly, the arrangement 190 of MlaF clearly resembles the pre-translocation state of MalK (Fig. 2D). This suggests that we 191 have trapped a similar conformation of the abMlaBDEF complex. It is possible that MlaD and/or 192 MlaF, for which there are no equivalent in other ABC transporters,stabilizes this conformation. 193 Alternatively, it is possible that the presence of detergents, which were present to solubilize the 194 complex, mimics the natural ligand in the transporter's active site. Finally, the transmembrane 195 (TM) region of the map is well resolved, and density for the transmembrane (TM) helices can be 196 clearly identified. We therefore modeled abMlaE, using an evolution restraints-derived structural 197 model of ecMlaE (21) as a template, and fitted the obtained coordinates in the corresponding 198 region of the map, with the orientation corresponding to the predicted topology. The resulting 199 MlaE dimer model (Fig. 2D) fits well to the EM map density (Fig. S5C), and clearly corresponds 200 to a closed transporter, with no solvent channel between the subunits. Interestingly, we also 201 noted clear density for three TM helices that likely correspond to the MlaD N-terminal helices 202 ( Fig. 3A). However, they lacked continuity, and we observed that only two form a direct 203 interaction with MlaE. It is possible that this is due to heterogeneity in the orientation of MlaD 204 relative to the rest of the complex. To verify this, we performed further 2D classification of the 205 particles used for reconstruction (Fig. 3B), which revealed a range of positions for the MlaD 206 region relative to the rest of the complex. We therefore performed further 3D classification 207 leading to a smaller dataset of ~ 8,000 particles. This produced a structure of lower resolution (~ 208 11.5 Å) but with the six MlaD N-terminal TM helices clearly visible (Fig. 3B). While the 209 periplasmic domain possesses 6-fold symmetry, the TM domains of MlaD do not appear 210 symmetrical, with two forming close contacts with the density attributed to MlaE while the other 211 four do not appear to contact any other proteins. This observation likely explains the asymmetry 212 of contacts between the dimeric MlaE and the hexameric MlaD. A higher-resolution structure 213 will be required to determine if additional contacts are formed between the outward-facing loops 214 of MlaE and the periplasmic domain of MlaD. 215 ΔmlaC A. baumannii had a dramatically decreased abundance of all major phospholipid species 247 in the OM compared to wild type. (Fig. 4A and Fig. S7). 248 To better analyze the differences in membrane GPL, we quantified GPL by normal phase 249 liquid-chromatography collision-induced-dissociation mass spectrometry (LC-MS/MS). We 250 quantified the ratio of individual GPL within each membrane by normalizing to an internal 251 standard of known quantity. We then normalized the quantified GPL to the protein content of 252 isolated IM and OM. Quantitative LC-MS/MS confirmed the overall reduction in outer 253 membrane GPLs observed by ESI-MS and TLC, with the reduced levels observable across 254 multiple GPL species for ΔmlaC mutants relative to wild type (Fig. 4B). Therefore, mutations in 255 the components of the Mla system result in a decrease in OM GPL, whereas the retrograde 256 transport hypothesis would predict an increase in OM GPL. Therefore, these results instead 257 suggest a possible role for Mla in outward GPL trafficking. 258 259

Mla mutants demonstrate an accumulation of newly synthesized GPL in the IM 260
The overall decrease in outer membrane glycerophospholipids of A. baumannii mla 261 mutants suggests that either the Mla system is functioning to deliver GPLs from the inner 262 membrane to the outer membrane, or alternatively, mutations in the Mla system may disrupt the 263 outer membrane in a manner that leads to the activation of outer membrane phospholipases, 264 which then degrade GPL. Work performed on the Mla system in E.coli has demonstrated that 265 disruption of genes in the Mla pathway results in activation of both the OM acyl-transferase 266 PagP, which cleaves a palmitate moiety from GPL and transfers it to LPS and PG , creating a 267 hepta-acylated LPS molecule and palmitoyl-PG and the OM phospholipase PldA (12,28). A. 268 baumannii has no known PagP enzyme but similar activity of the multiple predicted OM 269 phospholipases could account for the reduction in OM GPL as observed by TLC and quantitative 270 mass spectrometry. Therefore, we designed a mass spectrometry-based assay to study 271 intermembrane GPL transport using 13 C stable isotope labeling (Fig. S8A), to better analyze the 272 directionality of GPL transport by the Mla system between the bacterial membranes,. When 273 grown in culture with sodium acetate as the sole carbon source, many bacteria directly synthesize 274 acetyl-CoA using the conserved enzyme acetyl-CoA synthase (30). Acetyl CoA, the precursor 275 metabolite for fatty acid biosynthesis, is first converted to malonyl-CoA and enters the FasII 276 (fatty acid biosynthesis) pathway that supplies endogenously synthesized fatty acids to 277 macromolecules such as lipopolysaccharides, phospholipids, lipoproteins, and lipid-containing 278 metabolites. By growing cultures in unlabeled acetate then "pulsing" with 2-13 C acetate and 279 analyzing separated membrane fractions from set time points, we can observe the flow of newly 280 synthesized GPLs between the IM and OM of A. baumannii (Fig. S8B) (26). 281 Upon introducing the 2-13 C acetate as the sole carbon source, 13 C-labeled GPL were 282 immediately synthesized in the bacterial cytoplasm. We reasoned that continued growth in 13 C 283 acetate should result in a mixed pool of unlabeled and labeled IM GPL molecules. As the GPL 284 are then fluxed from the IM to the OM, the likelihood that an individual GPL molecule is 285 transported is directly proportional to the ratio of labeled to unlabeled GPL in the IM pool. As 286 the bacteria continue to grow in 13 C acetate, the ratio of labeled to unlabeled GPL in the IM will 287 gradually increase as new GPL are synthesized and inserted in the IM. As such, the likelihood of 288 transporting labeled GPL to the OM will also increase. A comparison of the ratios of labeled to 289 unlabeled GPL in the IM and OM will thus reflect the efficiency of transport between the 290 membranes, and analysis of transport in wild type A. baumannii will establish reference for 291 transport efficiency with which to compare our mutants. Additionally, OM phospholipases, some 292 of which may be activated upon membrane damage (31), will not distinguish between labeled 293 and unlabeled GPL and therefore will not affect the ratio of labeled to unlabeled GPL obtained 294 from this assay. 295 Membrane separation and analysis of wild type A. baumannii revealed near-identical 296 rates-of-change between the two membranes in ratios of 13 C-labeled to unlabeled GPLs,297 indicating that newly synthesized GPLs are transported and inserted into the OM at a rate 298 equivalent to their rate of synthesis and assembly within the IM. Furthermore, the ratios of 299 labeled to unlabeled GPLs were nearly equal in the IM compared to the OM at the time points 300 evaluated indicating that GPL transport likely occurs rapidly, consistent with earlier pulse-chase 301 experiments performed in E. coli that estimate the half-life of translocation of various GPLs at 302 between 0.5 and 2.8 min (32). By contrast, mutants in the Mla system accumulate newly 303 synthesized GPLs in their IM at a greater rate than occurs in the OM as evidenced by the 304 increasing disparity in the ratio of labeled to unlabeled GPLs between the IM and OM over time 305 ( Fig 5A). The discrepancy in ratios of labeled to unlabeled GPLs between the IM and OM of 306 ∆mlaF is apparent for PG and PE of varying acyl chain lengths corresponding to the most 307 naturally abundant species C16:0/C16:0, C18:1/C18:1, or C16:0/C18:1 (Table S2). Further, the 308 effects of MlaF K55L expression on GPL trafficking were similar to what was observed in the 309 ∆mlaF strain (Fig. 5B). Therefore, ATP hydrolysis by MlaF appears to be a requirement for 310 extraction of these GPLs from the IM of A. baumannii for subsequent transport to the OM. 311 To better characterize the role of the periplasmic substrate binding component MlaC, we 312 performed similar stable isotope pulse experiments to observe the flow of newly synthesized 313 GPLs in the ∆mlaC strains. Stable isotope experiments on ∆mlaC mutants reveal IM 314 accumulation of newly synthesized GPLs similar to the result in ∆mlaF mutants (Fig. S9A), 315 indicating that in the absence of the periplasmic component GPLs are not efficiently removed 316 from the IM by the remainder of the Mla system. We also sought to characterize the potential 317 role of the putative OM-lipoprotein MlaA, which has been implicated as a component of the Mla 318 system in E. coli. A chromosomal deletion strain of mlaA was created by allelic exchange, and 319 complemented by expression of MlaA from a pMMB67EH-Kan plasmid. The results of the 320 stable isotope pulse experiments in the ∆mlaA strain revealed results consistent with those 321 obtained from ∆mlaC and ∆mlaF, in which the ratio of labeled to unlabeled GPL is consistently 322 higher in the inner membrane than the outer membrane after one hour of exposure to 13 C-acetate 323 ( Figure S9B  We performed a screen to identify A. baumannii proteins that are essential for its OM 329 barrier that led to the identification of an ABC transport system whose ATPase activity maintains 330 OM barrier function. IM and periplasmic components of this system can be purified, bind GPLs, 331 and assemble into a defined protein complex with significant symmetry, indicating that this 332 system could function to transport GPLs from the IM to the OM. Consistent with the possibility 333 that Mla functions as an anterograde transporter, the OM of mutants show an overall reduction of 334 GPL along with an excess accumulation of newly synthesized GPL on the IM. Therefore, these 335 results lead us to propose that the function of the A. baumannii Mla system is the trafficking of 336 GPL from the IM, across the periplasm, for delivery to the outer membrane (Fig. 6) In this work, we designed a method to monitor lipid transport between Gram-negative 359 bacterial membranes using stable 13 C isotope labeling. Our results using this assay are consistent 360 with the Mla system functioning as an anterograde GPL transporter, however they do not 361 exclude the possibility of a dual role for Mla components in the maintenance of OM lipid 362 asymmetry. Previous work performed on the orthologous Mla system in E.coli has been 363 interpreted to suggest that the function of the system is to remove GPL from the outer leaflet of 364 the OM for retrograde transport back into the cytoplasm based on the observation that E.coli mla 365 mutants likely contain GPLs on the outer leaflet of the OM. (12,36). This proposed function was 366 inferred from the observation that gene deletions resulted in an increased activation of the OM-367 phospholipase enzymes PagP and OMPLA, suggesting an increased amount of GPL in the outer 368 leaflet of the OM (12). The interpretation of retrograde transport function was also based on the 369 existence of an orthologous system in plant chloroplasts that transports lipids from the 370 endoplasmic reticulum (ER) into the organelle. Many plants require this retrograde transport 371 function because certain lipids in the chloroplast thylakoid membrane derive from GPL 372 originating in the ER (37). However, since Gram-negative bacteria synthesize GPL within the 373 IM, retrograde transport of GPL would only be necessary for the recycling of GPL mislocalized 374 to the OM outer leaflet. Although this is a reasonable inference based on data available at the 375 time, we would point out that the directionality of transport by the E. coli Mla system had not 376 been thoroughly probed experimentally using membrane analysis or with a functional assay of 377 the type performed here. It is conceivable that the import function of the orthologous chloroplast 378 TGD system is a result of adaptation to the intracellular environment, the system in this case 379 having evolved to aid in the transfer of GPL from the nearby ER to the chloroplast. Furthermore, 380 while it is possible that the Mla system in E. coli serves a different primary function than in A. 381 baumannii, we demonstrate here that both complexes possess a similar architecture, pointing to a 382 conserved function. The outer membrane defect phenotypes observed in E. coli mla mutants 383 might also be explained by a disruption of OM structure stemming from decreased 384 concentrations of OM GPL, leading to activation of the PagP enzyme. It is well established that 385 for E. coli, GPL displacement to the OM outer leaflet and subsequent activation of these 386 enzymes reflects OM instability and can be achieved by chemical disruption of the bilayer (27-387 29). It may be the case that the OM of E. coli mla mutants contain GPL in the outer leaflet, but 388 the possibility remains that OM GPL can flip into the outer leaflet under conditions of OM 389 damage resulting from an imbalance of LPS-to-GPL ratios, along with perhaps the 390 corresponding disruption of OM proteins. However, final determination of the directionality of 391 GPL transport by the Mla system in E.coli and other organisms will require intermembrane 392 transport studies similar to what has been done here for A. baumannii, along with studies similar 393 to those performed for the Lpt LPS transport system for which molecular transfer of LPS from 394 molecule to molecule of the Lpt system is functionally defined. 395 The gene for MlaA, the proposed OM component, is at a different chromosomal location 396 from the remainder of the mla operon. Recent structural studies on MlaA have revealed that 397 MlaA forms a ring-shaped structure localized the inner leaflet of the OM, and have shown it to 398 form stable complexes with the outer membrane proteins OmpF and OmpC (38). The proposed 399 structure of MlaA in the OM supports the argument that MlaA is involved in removal of GPL 400 from the outer leaflet, and it is suggested that GPL from the outer leaflet travel through a pore in 401 MlaA where they are received by MlaC, yet our data reveals that A. baumannii ∆mlaA mutants 402 are defective in delivery of GPL from the IM to the OM. These data can be reconciled by a 403 model in which MlaA functions both to remove mislocalized GPL from the outer leaflet of the 404 OM, and additionally serves to facilitate delivery of GPL to the OM by MlaC, perhaps by 405 enabling MlaC localization to the surface of the inner leaflet. By this model, mutations in MlaA 406 will be phenotypically similar to mutations in other components of the Mla system, and we 407 would expect to observe a decreased rate of anterograde GPL transport. We would here point out 408 that while previous work has implicated the Mla system in the maintenance of OM lipid 409 asymmetry through observation of increased activity of PagP, the role of the MlaFEDB complex 410 and MlaC in retrograde GPL transport has previously only been inferred from homology to the 411 chloroplast TGD system. It is established that cellular mechanisms exist in Gram-negative 412 bacteria to resist stressful conditions that lead to OM disruption. For example, OM 413 phospholipase enzymes, such as PldA, are activated under conditions of membrane stress to 414 digest GPL in the outer leaflet of the OM, as high levels of GPL in the outer leaflet destabilize 415 the OM barrier function. The model of retrograde GPL transport by the Mla system proposes that 416 growing cells expend cellular energy in the form of ATP in order to transport undigested GPL 417 from the OM, across the periplasm, and back into the IM, at which point some of those same 418 molecules will be transported back to the OM by an unknown mechanism. However, the 419 available data points most clearly to a model of anterograde GPL transport by MlaFEDB and 420 MlaC, facilitated in some way by MlaA. 421 The first three genes of the mla operon -comprising an ATPase, permease, and substrate-422 binding components of the ABC transporter complex -are conserved in Mycobacteria spp, 423 Actinobacteria, and chloroplasts, while the entire five-gene operon appears to be conserved in 424 Gram-negative bacteria (39). Given the conservation of the system across Gram-negative 425 species, our results may shed light on a generalized mechanism contributing to OM biogenesis. 426 Additionally, we have here demonstrated that the function of this ABC transport system is 427 crucial for maintaining the integrity of the A. baumannii OM. The fact that mla mutations are 428 tolerated, and that levels of OM GPL are reduced but not abolished, suggests the intriguing 429 possibility of additional undiscovered mechanisms of GPL delivery to the OM. Also of interest is 430 the potential role of the increased exopolysaccharide observed upon disruption of the Mla 431 system. It is possible this exopolysaccharide plays a partially compensatory role in A. baumannii 432 resulting from decreased OM GPL, given that recent work has shown that A. baumannii 433 exopolysaccharides can contribute to antibiotic resistance, likely through improved barrier 434 function (40). 435 The progression towards a more complete understanding of intermembrane GPL 436 transport and OM barrier function should ultimately have relevance in the development of novel 437 drug targets to undermine emerging antibiotic resistance in Gram-negative pathogens. The 438 emergence of antibiotic resistant Gram-negative bacteria for which few or no antibiotics are 439 available therapeutically is an important medical concern. This issue is typified by current

MIC measurements: 496
MICs were determined in 96-well microtiter plates using a standard two-fold broth dilution 497 method of antibiotics in LB broth. The wells were inoculated with 10 4 bacteria per well, to a 498 final well volume of 100 µL, and plates were incubated at 37 °C with shaking unless stated 499 otherwise. Experiments were performed thrice using two technical replicates per experiment. 500 MICs were interpreted as the lowest antibiotic concentration for which the average OD 600 across 501 replicates was less than 50% of the average OD 600 measurement without antibiotic. Purified Mla complex at ~1 mg/ml was applied to glow-discharged holey grids, blotted for 6 s, 525 and plunged in liquid ethane using a Vitrobot (FEI). Images were acquired on a FEI Tecnai G2 526 F20 200 kV Cryo-TEM equipped with a Gatan K-2 Summit Direct Electron Detector camera 527 with a pixel size of 1.26 Å/pixel. 500 micrographs were collected using Leginon (47) spanning a 528 defocus range of -1 to -2 µm. 529 530 Movie frames were aligned with MotionCorr2 (48) and the defocus parameters were estimated 531 with CTFFIND4 (49). 333 high-quality micrographs were selected by manual inspection, from 532 which ~55,000 particles were picked with DOG in Appion (50). Particle stacks were generated in 533 Appion using a box size of 200 pixels. Several successive rounds of 2D and 3D classification 534 were performed in Relion 2 (51, 52) using an initial model generated by Common Lines in 535 EMAN2 (53) leading to a final stack of ~ 14,000 particles for 3D structure refinement in Relion. 536 537

Structure modeling and docking in the EM density: 538
The structures of MlaB and MlaF were modeled using the threading server Phyre (54) based on 539 the structures of the anti-sigma factor antagonist tm1081 (PDB ID 3F43, 18% sequence identity 540 to MlaB) and the ABC ATPase ABC2 (PDB ID 1OXT, 36% sequence identity to MlaF) 541 respectively. Two copies of each structural model were positioned in their putative location 542 within the EM map using Chimera (55) and their position was optimized using the Fit to EM 543 map option. The abMlaD and abMlaE structures were modeled on ecMlaD and ecMlaE 544 structural models deposited in the Gremelin database (21), using Modeller. For abMlaE, the 545 hexamer was modelled with Rosetta (56) as described previously (4). 546

Membrane Isolation and Separation 547
Cells were resuspended in 20 mL of 0.5 M sucrose, 10 mM Tris pH 7.8, 75 µg freshly prepared 548 lysozyme (Roche 10837059001), and 20 mL of 0.5 mM EDTA, and kept on ice with gentle 549 stirring for 20 min. Samples were homogenized (Avestin EmulsiFlex-C3) and spun down at 550 17,000 g for 10 min to removed un-lysed cells prior to ultracentrifugation. Membranes were spun 551 down using a Ti45 Beckman rotor at 100,000 g for 1 h and then added to the top of a sucrose 552 gradient. IM and OM were separated by 18-hour ultracentrifugation using a SW-41 rotor in a 553 Beckman Coulter Optima L90X ultracentrifuge. 554

G G G A G K V A A G S S S A E E K A P A S T D S S A Q P S F V E .
1 2 0 1 3 0 1 4 0 1 5 0 1 6 0 1 7 0 1 8 0 1 9 0 2 0 0 2 1 0 2 2 0 α1 β1 β2 β3 β4 β5 β6 β7 β8 α2          (2) the likelihood that a given GPL that is trafficked from the IM to the OM will be labeled is proportional to the ratio of labeled to unlabeled GPL in the IM;

. R Y N H I P D T S A R A N V F . E D P D M G T S I D G I Q D K D Q Y K S G G D G N A D T P A T P T A V A P T P V E N G S L P A T Q H P
(3) a comparison of the ratios of labeled to unlabeled GPL in the inner and outer membranes will therefore reflect the efficiency of GPL transport.