A dual substrate-accessing mechanism of a major facilitator superfamily protein facilitates lysophospholipid flipping across the cell membrane

Lysophospholipid transporter (LplT) is a member of the major facilitator superfamily present in many Gram-negative bacteria. LplT catalyzes flipping of lysophospholipids (LPLs) across the bacterial inner membrane, playing an important role in bacterial membrane homeostasis. We previously reported that LplT promotes both uptake of exogenous LPLs and intramembranous LPL flipping across the bilayer. To gain mechanistic insight into this dual LPL-flipping activity, here we implemented a combination of computational approaches and LPL transport analyses to study LPL binding of and translocation by LplT. Our results suggest that LplT translocates LPLs through an elongated cavity exhibiting an extremely asymmetric polarity. We found that two D(E)N motifs form a head group–binding site, in which the carboxylate group of Asp-30 is important for LPL head group recognition. Substitutions of residues in the head group–binding site disrupted both LPL uptake and flipping activities. However, alteration of hydrophobic residues on the interface between the N- and C-terminal domains impaired LPL flipping specifically, resulting in LPLs accumulation in the membrane, but LPL uptake remained active. These results suggest a dual substrate-accessing mechanism, in which LplT recruits LPLs to its substrate-binding site via two routes, either from its extracellular entry or through a membrane-embedded groove between transmembrane helices, and then moves them toward the inner membrane leaflet. This LPL-flipping mechanism is likely conserved in many bacterial species, and our findings illustrate how LplT adjusts the major facilitator superfamily translocation pathway to perform its versatile lipid homeostatic functions.

The major facilitator superfamily (MFS) 4 is the largest known secondary transporter protein family and consists of several hundred members in any cell type (1). MFS proteins transport a great variety of substrates including ions, sugars, amino acids, metabolites, nucleotides, peptides, and lipids. It is believed that MFS proteins share a common alternating-access mechanism, in which transporters undergo conformational changes to create alternating access to a centrally-located substrate-binding site from each side of the membrane (2). A rocker-switch model was supported by the structures of many MFS transporters, including sugar transporters LacY and GLUT1 (3,4). Of note, these structurally characterized MFS transporters mediate permeation of small polar substrates; i.e. they recruit substrates from the extracellular environment, move them across the hydrophobic membrane bilayer, and then release these solutes into the intracellular space. This translocation pathway is considered as a general working model for all MFS members (2).
Lysophospholipid transporter (LplT) belongs to the MFS family and is found in many Gram-negative bacteria. Distinct from other MFS members, LplT is a lipid flippase. LplT catalyzes flipping of lysophospholipids (LPLs) across the bacterial inner membrane (IM), playing an important role in bacterial membrane phospholipid homeostasis. In bacteria, LplT is functionally linked with de novo biosynthesis of the major outer membrane lipoprotein (Lpp). Generation of matured Lpp requires acyl-transfer from diacyl phospholipids to its N terminus, which releases LPL as a by-product into the outer leaflet of the IM (5) (reaction catalyzed by apolipoprotein N-acyltransferase Lnt, Fig. 1A). LPLs are nonbilayer forming lipids and may destabilize the lamellar membrane structure by inducing positive curvature (6). LplT catalyzes energy-independent flipping of LPLs from the outer leaflet to the inner leaflet of the IM (7) where they are acted upon the peripheral acyltransferase/acyl-acyl carrier protein synthetase Aas on the cytoplasmic surface for acylation to form their respective parent forms (7,8). Our recent study showed that mutational inactivation of LplT in Escherichia coli impairs the stability of the IM and lipid asymmetry of the outer membrane (OM) mediated by accumulation of LPLs in the IM (9). This intramembranous LPL-flipping activity is apparently different from the common MFS working model, suggesting that LplT utilizes a specific MFS transport mechanism for lipid flipping.
In addition to the intramembrane-flipping activity, our recent biochemical study also found that LplT catalyzes direct uptake of exogenous LPLs into OM-depleted spheroplasts (8,9) (Fig. 1A). This LPL influx activity appears to be independent of intramembranous flipping because spontaneous incorporation of LPLs into the membrane bilayer is very inefficient (8). How LplT promotes these two distinct transport modes, whether by sharing one translocation pathway or using two individual routes, remains unknown.
Unlike other known flippases, LplT maintains an extreme specificity for lyso versus diacyl forms. We found that spheroplasts expressing LplT from Klebsiella pneumoniae (LplT Kp ) transports LPLs including both lysophosphatidylethanolamine (LPE), lysophosphatidylglycerol (LPG), but is not inhibited by diacyl-PE or -PG (8). This acyl-chain selectivity may be critical for LplT to perform its specific LPL-flipping activity in the diacyl lipid-rich membrane bilayer.
Lysophospholipid transporter, namely MFSD2A, is also found in animals, although its sequence homology to bacterial LplT is rather poor. MFSD2A has been found to be essential for brain growth and function. MFSD2A transports exogenous lysophosphatidylcholine (LPC) across the brain-blood barrier as the major route for supplying the -3 fatty acid docosohexaenoic acid to the brains (10). Interestingly, dual activities have also been reported in MFSD2A. In addition to its LPL influx activity, MFSD2A also maintains the integrity of the brain-blood barrier by suppressing transcytosis in the central nervous system epithelial cells (11). Whether this membranemaintaining role is mediated by the LPC uptake or other activities of MFSD2A remains to be addressed (12).
No structure of LplT or MFSD2A is available to date. To understand the novel dual LPL uptake/flipping mechanism of LplT, we constructed a three-dimensional structural model of LplT Kp , docked the models with different LPL ligands, and studied its LPL uptake and intramembranous flipping activities using mutational analyses. Our results not only provide molecular insights into how LplT recognizes the chemical structure of the LPL substrates, but also reveal a new MFS transport mechanism for LPL translocation across the cell membrane.

Dual substrate transport activities of LplT Kp
The intramembranous LPL-flipping activity of LplT Kp was assessed by examining LPL contents in the IM. OM-depleted spheroplasts were generated from metabolically 32 P-labeled E. coli BL21(DE3) cells and then washed to remove any extracellular components carefully prior to lipid analysis using thinlayer chromatography (TLC). As shown in Fig. 1B and Table 1, no LPLs were detected in WT spheroplasts. In contrast, LPE and LPG were accumulated to 17 and 5% of the total phospholipid compositions, respectively, in ⌬lplT spheroplasts. This LPL accumulation was completely diminished in the presence of LplT Kp . The Lpp acylation reaction catalyzed by Lpp acyl apolipoprotein N-acyltransferase (Lnt) is the only known source to generate LPLs in the IM (5). To exclude any change of LPL production in the mutant strain, we analyzed Lpp acylation using Western blotting with a specific anti-Lpp antibody. PAP9502 is a conditional lnt-depletion E. coli strain in which endogenous expression of Lnt is controlled by an P BAD promoter (13). In the presence of glucose, the lack of Lnt resulted in an unacylated Lpp precursor in the cells, which migrated faster than its mature form on urea-denaturing gel (Fig. 1C). No Lpp precursor was detected in either ⌬lplT or ⌬aas strains and the matured form of Lpp was present in a similar level compared with WT. Therefore, it is most likely that the accumulation of LPE/LPG in ⌬lplT spheroplasts (Fig. 1B) represents the flipping activity of LplT Kp .
The extracellular LPL uptake activity of LplT Kp was measured by adding [ 32 P]LPE into OM-depleted spheroplasts generated from E. coli cells. Previously, we have demonstrated the LPL-transport activity of LplT in an E. coli aasϪ strain, indicating that LplT can catalyze LPL transport independent of Aas (9). To better characterize the transport activity of LplT Kp , the transport assay was performed in the background of E. coli lplT-aas double deletions. As shown in Fig. 1D, [ 32 P]LPE was imported into spheroplasts in the presence of LplT Kp . In contrast, no uptake was detected in the control spheroplasts. Despite the fact that LPLs are detergent-like molecules, spontaneous incorporation of LPE into the bacterial membrane is negligible in the absence of LplT Kp at the assay conditions. We further confirmed it using inside-out vesicles (ISO) (Fig. 1D). Therefore, it is unlikely that extraneous LPLs access the translocation pathway of LplT Kp through the bilayer. Instead, they may directly access the pathway from its extracellular protein surface. Taken together, these results demonstrate that LplT Kp may utilize two distinct routes to perform each individual LPL uptake or flipping activity (Fig. 1A).

Construction of an LplT Kp structural model
To gain structural insights into the novel dual LPL-flipping activity of LplT, we took advantage of the conserved membrane topology and protein architecture of MFS proteins to construct a structural homology model of LplT Kp . Despite the determination of nearly 100 MFS protein structures, the low sequence identity (Ͻ25%) between LplT and other known MFS protein structures is a major hurdle to structural modeling of LplT. Attempts to search for a suitable structural template for LplT Kp using the PSI-BLAST program (14) did not yield any hits. Thus, we made use of four automatic structure prediction programs including HHpred (15), Phyre2 (16), RaptorX (17), and I-TASSER (18) to identify remote structural homologs and create initial structural models of LplT Kp . We also carried out homology modeling of LplT Kp independent of these structure prediction programs using MODELLER v9.14. Sampling of the conformational space was enhanced in our homology modeling protocol by using diverse template-target alignments generated Lipid flipping mechanism of MFS protein by different sequence alignment methods (Fig. S1, 1-25). By such, a large pool containing 15,000 LplT Kp models was created, from which 600 candidates were selected for substrate docking tests using the DOCK 3.6 program (19).
LplT Kp has a unique substrate selectivity of the LPL head group; i.e. it transports LPE and LPG, but not LPC and lysophosphatidic acid (LPA) (8), despite the fact that both LPE and LPC are zwitterionic lipids and share a similar quaternary amine head group. We hypothesized that an ideal LplT Kp struc-tural model should be able to distinguish its native substrates (LPE and LPG) from nonsubstrates (LPC and LPA). Considering the fact that C16 and C18 acyl chains are present predominantly (Ͼ95%) in E. coli membrane phospholipids (20), the docking screening was performed with C18:1 LPL substrates. Both inward-and outward-open templates were used for modeling. However, we found that models generated based on outward-open templates were much better at distinguishing between LPE/LPG and LPC compared with those based on inward-open templates. Of 300 inward-open models, only 51 models were able to distinguish LPE/LPG from LPC, whereas 72 outward-open models of 300 were able to do the same. Models based on outward-open templates showed clearer demarcation between LPE/LPG and LPC (based on docking scores) compared with inward-open templatebased models. We found that a MODELLER-generated model using the structure of a putative MFS transporter YajR Ec protein from E. coli (PDB code 3WDO) as the template and template-target alignment created using PRALINE pairwise alignment exhibited docking scores, which qualitatively matched well with the known substrate selectivity of LplT Kp : LPE Ϸ LPG Ͼ LPA Ͼ Ͼ LPC ( Fig. 2A). This model was considered to be the most accurate homology model of LplT Kp and used for further mechanistic studies.

Lipid flipping mechanism of MFS protein Overview of the LplT Kp model
The LplT Kp model consists of 12 canonical transmembrane helices (TM1-12) and shows a typical MFS protein overall conformation (Fig. 3, A-C). TM1-6 and TM7-12 are arranged into two 6-helix bundles to assemble the N-and C-domains with a 2-fold pseudo symmetry. The two domains are linked by a hydrophilic intracellular loop region.
Despite LplT Kp and YajR Ec sharing a very low protein sequence identity (ϳ11%) (Fig. S1, 23), the LplT Kp model exhibits a similar conformation as that of the YajR Ec structure with a r.m.s. deviation value of 3.2 Å (Fig. 4B). Both structures also exhibit a remarkable similarity in the distribution of charged residues, particularly those basic residues congregated on the intracellular protein surface (Fig. 4A). Motif A is a fingerprint sequence found in the N-domain of most MFS proteins and assembles a conserved salt-bridge interaction network thought to be important for the outward-open conformation (21). In the YajR Ec structure, Motif A residues Gly-69 TM2 , Asp-73 TM2 , and Arg-77 IL2-3 , and Arg-77 IL2-3 interact with Asp-73 TM2 and Asp-126 TM4 (Fig. 4B). A similar salt-bridge network is formed by the counterpart residues Gly-70 TM2 , Asp-74 TM2 , Lys-78 IL2-3 , and Glu-126 TM4 in the LplT Kp model. Charged residues also form conserved salt-bridge interactions in the C-domain of MFS proteins. These interactions were found at the structurally equivalent positions in the LplT Kp model, although some interacting residues appear to switch their positions with their partners compared with YajR Ec (Fig. 4B); i.e. Glu-272 TM8 and Lys-329 TM10 forming a salt-bridge in YajR Ec are substituted by Lys-275 TM8 and Glu-332 TM10 in LplT Kp , respectively. Similarly, two acidic residues Glu-211 IL6 -7 and Glu-330 IL10 -11 that interact with Arg-213 TM7 in YajR Ec are substituted by Arg-214 IL6 -7 and Arg-333 IL10 -11 to interact with Glu-217 TM7 (ϩ1) in LplT Kp . Taken together, the presence of these conserved salt-bridge networks serves as additional validation for the LplT Kp model.

LPL substrate docking conformations
In the LplT Kp model, a large and elongated central cavity is found on the interface between the N-and C-domains, creating a substrate translocation pathway (Fig. 3A). The model has an apparently outward-open conformation, in which the pathway is opened toward the periplasmic surface and ends in the middle of the membrane (Fig. 5). The highest ranked docking poses of LPE and LPG show that the substrates are bound in a similar head-to-tail orientation from the extracellular to the intracellular side in the central cavity (Fig. 5, A and B). Binding of LPE/LPG is likely facilitated by complementation of the asymmetric polarity in the cavity as illustrated by the electrostatic potential map (Fig. 5E). Charged/ polar residues are predominantly distributed toward the

Lipid flipping mechanism of MFS protein
center of the translocation pathway, whereas hydrophobic residues are found toward the extracellular side. Of note, these cavity-lining residues are conserved in the LplT family (Fig. 6, A and B).

Head group-binding site
Both LPE and LPG were favorably bound in a fishhook-like conformation in the central cavity (Fig. 6, C and D). This substrate conformation is apparently stabilized by two symmetric    (40). Regions having positive (blue) and negative (red) potentials surround the negatively-charged phosphate group and positively-charged ethanolamine head group, respectively, whereas neutral (white) regions surround the acyl tail.

Lipid flipping mechanism of MFS protein
The importance of these head group interactions for LPL transport activity was demonstrated using in vivo LPE acylation assays (Fig. 7, A and B). In the assay, LplT Kp WT assimilated exogenous [ 32 P]LPE into OM-depleted spheroplasts and 40% of added LPE was converted to diacyl-PE by endogenous Aas in 30 min. Remarkably, mutation of Asp-30 to asparagine completely abrogated the LPE acylation activity, whereas substitution of Asn-352 with a cysteine residue drastically reduced the activity by Ͼ80%. Substantial reductions (30 -40%) were also observed in the mutants of E351Q and K120C. The loss of LPE acylation activity is clearly attributed to the changes in LplT Kp -mediated LPL uptake given the facts that (i) disrupting the mutant spheroplasts using Triton X-100 detergent retrieved the full Aas activities (Fig. 7, A and B); (ii) the mutant proteins were Lipid flipping mechanism of MFS protein present in spheroplasts in similar levels compared with WT as examined using immunoblotting (Fig. 7C).
To help understand the role of these head group interacting residues, we also tested substrate docking to in silico LplT Kp mutants. D30N and K120C mutants were generated in silico using University of California San Francisco, Chimera rotamers module (22,23). Both mutant models exhibited significantly reduced ability to bind to 18:1 LPE and LPG based on their docking scores (Fig. 2B). Taken together, these data suggest that LplT Kp utilizes the central cavity for LPL substrate binding.

Molecular determinant of head group discrimination
In the docked models, the carboxylate group of Asp-30 appears to be in the critical position to interact with the head group moiety of LPE or LPG and may contribute to the LPL head group selectivity of LplT. To test this hypothesis, we compared the transport activities of LplT Kp WT and the D30N mutant using [ 32 P]LPE and LPG. The results showed that LplT Kp WT transports LPE and LPG with similar kinetic parameters: 2.3 Ϯ 0.1 M (K m ) and 98.4 Ϯ 7.3 nmol/g dry weight/h (V max ) for LPE; 2.2 Ϯ 0.2 M (K m ) and 80.2 Ϯ 6.4 nmol/g dry weight/h (V max ) for LPG. Consistent to the acylation assay, the mutation of D30N significantly reduced the transport activities for both LPE and LPG. Kinetic analysis showed reductions of both K m and V max for LPE by 4-or 2-fold to 9.5 Ϯ 1.0 M (K m ) and 64.4 Ϯ 5.6 nmol/g dry weight/h (V max ), suggesting that the carboxylate group of Asp-30 is involved in both substrate-binding and transport. Similar changes were also observed in transport of LPG (V max ϭ 56.4 Ϯ 4.3 nmol/g dry weight/h, K m ϭ 6.9 Ϯ 0.7 M). These results support our model prediction that both substrates share similar binding conformations in the central cavity (Fig. 6, C and D). In fact, the importance of head group interaction with Asp-30 was also demonstrated in LPA docking; i.e. C18:1 LPA was able to dock in the binding site in a fishing-hook pose similar to that of LPE and LPG (Fig. 5C). But the lack of a head group moiety led to increased free energy based on the docking score ( Fig. 2A).
LplT cannot transport LPC (8). The head group of LPC also carries a cationic amine group, which could potentially interact with Asp-30. However, the most energetically favorable docked pose for 18:1 LPC precludes such an interaction. LPC is positioned in the opposite orientation compared with LPE and LPG in the central cavity, i.e. a head-to-tail orientation along the intracellular-to-extracellular direction (Fig. 5D). For C18:1 LPC, at a total volume of ϳ763 Å 3 , is bulkier than C18:1 LPE (ϳ677 Å 3 ) and C18:1 LPG (ϳ704 Å 3 ), which could prevent access to the head group-binding site in the middle of the membrane. In fact, the reason for the exclusion of the LPC head group from the polar region of the central cavity becomes

Lipid flipping mechanism of MFS protein
apparent when an LPC molecule was superposed onto the docked LPE molecule in the LplT Kp model (Fig. 6D). The bulky trimethyl moiety of LPC experiences severe van der Waals clashes with Asp-30, preventing its direct interaction with the head group of LPC.

LPL acyl tail interaction
The central cavity toward the extracellular side is primarily lined by aromatic and aliphatic residues. Three residues from TM2 including Leu-34 TM2 , Phe-35 TM2 , and Leu-38 TM2 occur within 4 Å of the acyl chain of the docked 18:1 LPE/LPG (Fig.  8A) and could represent important nonpolar interactions for LPL recognition and transport. We found that mutations of L34F, F35N, and L38F reduced LPE acylation activity modestly by 20 -40% (Fig. 7, A and B). We hypothesized that these aliphatic residues provide a favorable "greasy patch" to accommodate the acyl chain of LPL during the transport cycle. In fact, a similar structural feature was found to be crucial for the transport activity of MFSD2A (24). Hydrophobic interactions with the monoacyl chain may facilitate LPL flipping in general.
An interesting question is whether the large monoacyl fatty chain also plays any role in maintaining the substrate selectivity. To gain any molecular insight into this question, we performed additional docking studies using shorter chain substrates including 16:1 and 14:1 LPE, LPG, and LPC molecules.
The docking scores for the best scoring poses are presented in Fig. 2C. From the docking scores it is evident that, with a decrease in acyl chain length, the docking scores (kcal/mol) improve for docked ligands (especially for LPC), whereas the discriminatory capability of the LplT Kp model seems to diminish when the acyl chain length is reduced to 14. Examination of the docking poses show that all six shorter acyl chain LPLs are snuggly buried within the central pocket with their acyl chains pointing toward the extracellular side (Fig. S2). The docking conformations of 16:1 and 14:1 LPC exhibit an orientation that is reversed compared with that of the 18:1 LPC. Despite the reversed orientation, analysis of intramolecular contacts indicates that the trimethyl moiety of 16:1 LPC still encounters severe van der Waals clashes with the side chain atoms of polar residues within the pocket. Such clashes are much less with 14:1 LPC. These observations suggest that the nature of both the head group and acyl chain could potentially contribute to the substrate selectivity of LplT.
A positively charged residue Arg-236 TM7 was found in the largely apolar region of the central cavity (Fig. 8A). Mutation of R236M led to nearly 40% reduction of the LplT Kp activity based on the LPL acylation assay (Fig. 7). Given its strategic location, we anticipate that the basic side chain of Arg-236 is involved in the early phases of LplT Kp 's interaction with the head group of

Lipid flipping mechanism of MFS protein
LPLs. Also, its potential role in forming an extracellular "gate" with Asp-30/Asn-31, similar to the one observed in E. coli GlpT (25), cannot be ruled out.

Substrate accessing pathway for LPL flipping
The central cavity is the only substrate translocation pathway found in the LplT Kp model, thus it may also be used for intramembranous LPL flipping. The five residues in the head group-binding site appeared to be essential for the flipping activity; i.e. spheroplasts of D30N, K120C, N31C, E351Q, or N352C accumulated both LPE and LPG to the ⌬lplT level (Fig.  8C, Table 1). These results suggest that LplT Kp utilizes the same substrate-binding site for both intramembranous LPL flipping and extracellular LPL uptake.
One important question is how LPL substrates gain access to the central translocation pathway from the membrane bilayer.
In the LplT Kp model, the outward-open conformation creates two V-shaped grooves between TMs 5 and 8 and between TMs 2 and 11 (Fig. 8B). These two grooves open toward the outer leaflet of the membrane and may serve as membrane entry points for LPL substrates. In support of this idea, we have found that the hydrophobic residues predicted for the acyl chain interaction on TM2 become more prominent for the LPL flipping activity compared with their modest roles in LPL uptake. Mutations of L34F, F35N, and L38F led to accumulation of ϳ15% PE and ϳ8% PG, similar to the levels observed in ⌬lplT (Fig. 8C, Table 1). Furthermore, we found that mutation of Ile-148 TM5 located in the middle of the groove between TMs 5 and 8 to a bulky phenylalanine caused a significant increase in LPL levels by Ͼ40% compared with WT (Fig. 8C, Table 1), whereas its LPL uptake activity remains fully active (Fig. 7). These results suggest hydrophobic interaction plays a critical role in mediating LPL intramembranous flipping.

Intracellular exit
In the outward-open model of LplT Kp , LPE/LPG are bound in the extracellular half of the translocation pathway and the pathway is apparently closed at the intracellular side of the molecule. Several polar residues including Gln-23 TM1 , Asn-137 TM5 , Asn-327 TM10 , Gln-331 TM10 , Gln-347 TM11 , and Asn-348 TM11 were found toward the intracellular side of the central pathway (Fig. 9A). Although these residues have no direct inter-actions with LPE/LPG, we asked whether these residues facilitate translocation of the substrate through the pathway. However, mutation of these residues to cysteine, respectively, yielded no effect on LPL transport activity (Fig. 9B) and no LPL accumulation was detected in N137C spheroplasts (Fig. 7). Although any large structural rearrangements in the pathway cannot be ruled out, our results suggest that LplT Kp utilizes a different exit than the central port to deliver LPLs to the inner leaflet of the membrane.

Discussion
Maintaining phospholipid homeostasis is important for the structure of cell membranes and the activity of membrane proteins (26,27). This process is facilitated by multiple mechanisms in which lipid flippases and scramblases play an essential role (28,29). LplT is unique among the known flippases due to its substrate specificity for LPLs (8). In this study, we proposed a new dual LPLflipping mechanism of MFS transporters. We took advantage of conserved MFS protein structures to create a structural model of LplT Kp . The model, supported by biochemical characterizations, suggests that LPL transport requires recognition of both the head group and fatty acyl chain. Direct interaction between the carboxylate group of Asp-30 and the head group moiety is important for substrate binding and selectivity (Fig. 7, A and B). In the model, LPE/LPG are stabilized in a fishhook-like conformation primarily by the two D(E)N motifs in the binding site (Fig. 6, C and D). This substrate-binding conformation lays the glycerol 3-phosphate backbone in the bottom of the binding cavity and inevitably places its sn-2 -OH group at an inaccessible position, precluding the sn-2 acyl chain of any diacyl phospholipids from penetrating the narrow central cavity.
The docking with different acyl chain substrates also suggest a role for the long acyl chain in maintaining the selectivity of LPE/LPG versus LPC of LplT. We found, in silico, that the discrimination of LPC is strictly maintained with C16 -C18 LPLs, but is quickly diminished when the chain length is reduced to 14 (Fig. 2C). It seems that the longer acyl chain enhances structural hindrance of the PC head group despite the nature of its flexibility. LPC is not synthesized in E. coli cells, but is present abundantly in body fluids (30). C16 -20 LPC are the most prominent LPC species generated by host endothelial lipase (30). Therefore, establishing the distinguishability among the

Lipid flipping mechanism of MFS protein
LPLs at the level of C16 -18 could be sufficient to prevent assimilation of exogenous LPC into the bacterial cells. Further testing of this hypothesis may help gain novel insights into the mechanism of substrate binding and selectivity of LplT.
LplT Kp catalyzes both exogenous LPL uptake and intramembranous flipping of LPLs across the bacterial IM. Our results strongly suggest that both events take place in the same translocation pathway between the N-and C-domains (Figs. 7, A and  B, and 8C). The ability of LplT Kp to assimilate LPLs from the extracellular space is remarkably efficient due to its high substrate binding affinity (micromolar range). LPLs exist as a mixture of micelles and monomeric forms in solution due to their high critical micelle concentrations (31). Based on the model, monomeric LPE/LPG molecules may access the central cavity directly in a diving-like pose. The residue Arg-236 that is located near the periplasmic entry region (Fig. 8A) could mediate the passage of the polar group through the apolar region toward the head group-binding site.
However, this extracellular entry is apparently inaccessible to LPLs in the membrane because LPLs generated in the outer leaflet of the IM stay in the bilayer as demonstrated in ⌬lplT spheroplasts (Fig. 1B) and the fact that no spontaneous incorporation of LPLs was observed in both spheroplasts and ISO control vesicles (Fig. 1D). The cavity-lining residues, responsible for the binding of LPL head group, are also essential for LPL-flipping activity (Fig. 8C), suggesting that membrane LPLs are positioned in the central cavity in a similar asymmetric orientation. According to the MFS rocker-switch mechanism, substrate binding in the cavity may induce a rigid body rotation between the N-and C-domains along TMs 5 and 11, resulting in two V-shaped openings on both sides of the translocation pathway (32). Mutation of Ile-148 in the groove between TMs 5 and 7-8 only impairs the LPL flipping activity without any interruption of LPL uptake (Fig. 7, A and B, and 8C), arguing for its specific role in mediating the membrane LPL entry rather than for general transporter operation. We envision that the V-shaped grooves between the N-and C-domains serve as a membrane-embedded entry for slender LPL accessing into the central-binding cavity from the outer leaflet (Fig. 8B). This hypothesis is further supported by mutation of hydrophobic residues on TM2. Mutation of L34F, F35N, or L38F completely abolished the LPL-flipping activity based on the levels of LPLs accumulated in spheroplasts (Fig. 8C). These residues are located on the periplasmic half of the pathway and adjacent to the V-shaped grooves (Fig. 8A). It is possible that they help orientate LPL substrates from the membrane toward the head group-binding site for translocation.
It has long been enigmatic how LPL substrates pass through their translocation pathway across the membrane. How LplT achieves alternating-access operation is still unclear because mutations near the cytoplasmic exit gave no effect on the transport activity (Fig. 9). Unlike other MFS transporters that move substrates through the translocation pathway, the major function of LplT is to orientate the LPL to the right position for acylation by Aas in the membrane. In the LplT Kp model, substrate bound in the elongated cavity arranges its head group to crossover the mid-line of the bilayer. It is possible that an exit is then created on the membrane interface to release substrates laterally into the inner leaflet of the membrane. Despite the fact that LplT and Aas may work independently (9), whether Aas is involved in LPL exiting from LplT is still unknown. Flipped LPLs may need to be acylated immediately by Aas to eliminate any membrane disruptive effect in the inner leaflet. In many bacteria including E. coli and K. pneumoniae, the genes of LplT and Aas are adjacently encoded in the same operon (7). However, in other Gram-negative microorganisms, LplT and Aas are predicted to form a fusion protein. Future investigation of any protein interaction between LplT and Aas may help to understand LPL substrate exiting from LplT.
LplT is an energy-independent transporter (7) and its transport activity should be driven by the concentration gradient of substrates across the bilayer. However, the extremely asymmetric polarity in the central translocation pathway may only support LPL influx. No significant role of those polar residues toward the cytoplasm in the pathway (Fig. 9) perhaps also prohibits LPL substrate binding from the cytoplasmic surface. Considering the facts that phospholipid degradation occurs mostly in the OM and the outer leaflet of the IM and no acyltransferase has been found in the periplasm (33), it may be unnecessary to maintain an LPL efflux activity in the bacterial membrane.
The functional role of this dual-substrate accessing mechanism in LplT-mediated bacterial membrane homeostasis is still unclear. Our recent study has showed that LplT/Aas maintain the structure and stability of both the IM and OM (9). The dual-substrate accessing mechanism may facilitate the functional versatility of LplT in the bacterial membrane envelope. In addition to LPLs generated from Lpp biosynthesis, LPLs can be produced from OM hydrolysis by secretory phospholipase A2 serving as a potent host anti-bacterial mechanism (33). We found that inactivation of LplT in E. coli greatly facilitates mammalian phospholipase A2-mediated attack (9). The dual-access mechanism may enable LplT to scavenge LPLs generated from multiple phospholipid degradation pathways (33). Whether this LPL transport mechanism is applicable to other LplT members requires further characterization. What is clear is that LplT has orthologues across multiple species. On the basis of the conservation of characteristic residues that interact with LPL substrates in the modeled structures, respectively, 150 other sequence homologs (pairwise sequence identity of 35.2-85.6% with LplT Kp ) from different prokaryotic species may now be classified as lysophospholipid transporters with a similar transporting mechanism (Fig. 6, A and B). In all of these sequences, the cavity-lining residues (Asp-30, Asn-31, Lys-120, Asn-137, Arg-236, Glu-351, Asn-352, Leu-34, Phe-35, and Leu-38, LplT Kp numbering) are highly conserved, as is the residue located in the middle of the groove between TMs 5 and 8 (Ile-148). Therefore, the transport mechanism demonstrated in LplT Kp may serve as a model to understand the role of LplT in lipid homeostasis.

Preparation of E. coli spheroplasts
Spheroplasts were generated from E. coli cells using a lysozyme-LiCl approach as we previously used (8). Prior to preparation, LplT Kp WT or mutant proteins with a His tag at the N terminus was expressed in E. coli BL21(DE3) strains using a pET vector. The mutations were introduced using a standard site-directed Lipid flipping mechanism of MFS protein mutagenesis method. E. coli cells were grown in LB broth at 37°C until an A 600 of 0.5. Protein expression was induced by adding 0.2 mM isopropyl ␤-D-1-thiogalactopyranoside for 2 h at 25°C. Cells were washed with a buffer containing 10 mM HEPES (pH 7.5), 0.75 M sucrose, 10 mM MgSO 4 , and 460 mM LiCl. After addition of 1 mg/ml of lysozyme, cell suspensions were ice-chilled, warmed to room temperature, and then incubated with gentle shaking at 30°C for 30 min. Intact spheroplasts were collected by centrifugation (3,000 ϫ g for 10 min) at room temperature and resuspended at 10 mg/ml of total protein concentration in the above buffer without LiCl for immediate use. Spheroplast formation and stability were thoroughly monitored nephelometrically by following the A 600 of 100 l of a spheroplast suspension in 2 ml of either water or a solution supplemented with 10 mM MgCl 2 , 0.75 M sucrose, respectively.

LPE acylation assay
LPE acylation assays were performed in spheroplasts generated from E. coli BL21(DE3) ⌬lplT mutant strains as previously described (8). Prior to assays, [ 32 P]LPE and -LPG were prepared by digestion of metabolically 32 P-labeled PE or PG using venom PLA 2 and then purified by thin-layer chromatography (TLC). [ 32 P]LPE (10 5 cpm for each assay) mixed with 10 M (final concentration) 18:1 LPE (Avanti lipids) was added into spheroplast suspensions for 30 min. The reactions were terminated by adding chloroform/methanol (1:2) to extract total lipids using the Bligh-Dyer method. Lipids were separated on Silica Gel G thinlayer plates and developed with chloroform/methanol/ammonia/H 2 O (60:37.5:1:3, v/v). The dried plate was exposed to a Storage Phosphor Screen overnight. PE and LPE were visualized and quantified using Molecular Imager FX. Stored images were processed and quantified using Quantity One software. LPE acylation activity is expressed as mole % of PE/(PE ϩ LPE).

Analysis of LPL contents in IM
LPL contents in the E. coli inner membrane was measured by TLC analysis of LPLs in spheroplasts. Cells were grown in the presence of 5 Ci/ml of [ 32 P]orthophosphate. Spheroplasts were generated using the same approach as described above. Spheroplasts were washed carefully three times with the buffer prior to lipid extraction and TLC analysis. The content of each individual lipid species is expressed as mole % of the total phospholipid composition.

Immunoblotting
The Lpp protein in E. coli was assayed by following a published protocol (13). Crude extracts from PAP9502 (BW25113 ybeX-(kan-rpoCter-paraB)-lnt), W3110 WT, ⌬lplT, and ⌬aas strains were prepared from cells grown in LB broth containing 0.2% arabinose or glucose for 8 generations at 37°C. The proteins were separated on urea SDS gel and developed with an anti-Lpp antibody. The expression of LplT Kp WT and variants were analyzed using an anti-His antibody.

LPL incorporation in inside-out vesicles
Inside-out membrane vesicles were prepared using a lowpressure homogenization approach described earlier (34). Briefly, cells suspended in cold Tris-HCl (50 mM, pH 8.0) were ruptured by single passage through a C3 homogenizer (Avestin) at 4,000 p.s.i. After removing cell debris, the supernatant was pelleted by ultracentrifugation at 40,000 rpm using a Ti45 rotor for 1 h. Membrane fractions were rinsed and homogenized in the same Tris-HCl buffer. Prior to assays, vesicles were diluted into a buffer containing 0.1 M HEPES (pH 7.5), 100 mM KCl, and 2 mM NADH to 0.13 mg/ml of total protein concentration. The reactions were triggered by adding 10 M [ 32 P]LPE (10 5 cpm) for the indicated times at 37°C. The reactions were terminated by filtration through a 0.22-m membrane on a Millipore filtration manifold and washed with the reaction buffer. The radioactivity in the membrane was counted using a liquid scintillation counter to calculate transport activity.

LPL transport assay in spheroplasts
Transport assays were carried out in spheroplasts generated from E. coli BL21(DE3) ⌬aas-lplT strains using a silicon oilbased approach as we described previously (8). Briefly, [ 32 P]LPE or LPG mixed with synthetic LPE (18:1) or LPG (18:1) were added to the spheroplasts at different concentrations for 30 min at 37°C. After incubation, 0.5-ml reactions were layered onto 0.15 ml of 22% perchloric acid and 0.50 ml of silicone oil in Eppendorf tubes and centrifuged at 14,000 ϫ g for 5 min at room temperature to separate spheroplasts from free substrates. The radioactivity in the perchloric acid phase was quantified by a liquid scintillation counter to calculate transport activity. Michaelis-Menten kinetic constants were analyzed using the GraphPad Prism software.

Three-dimensional structure prediction
The structural model of LplT Kp was constructed using a similar homology modeling approach previously used for the human sodium-dependent LPC transporter MFSD2A with modifications (24). The three steps in the modeling approach are as follows. 1) The initial templates were identified using four automated protein structure prediction programs HHpred (15), Phyre2 (16), RaptorX (17), and I-TASSER (18). The most commonly identified templates that were used by the structure prediction programs include PDB 3O7Q, 5AYN, 3WDO (outward-open conformation), and 4ZP0 and 1PW4 (inward-open conformation) (Table S2). We also included PDB 5AYO as one of the templates whose primary sequence is identical to that of PDB 5AYN but the solved structure is in the inward-open conformation. 2) In addition to the template-query alignments generated by the structure prediction programs, to further improve sampling, pairwise alignments between each of the template sequences and LplT Kp were generated using five different alignment approaches, namely pairwise, and profile-profile alignment using AlignMe (35) and MUSCLE (36), and pairwise alignment using PRALINE (37). Depending on the chosen template and alignment approach used, 30 different combinations were considered and 500 homology models were constructed for each template-alignment combination using the automodel class in MODELLER v9.14 (38). The modeling approach comprises structure optimization by conjugate gradients and refinement by molecular dynamics and simulated annealing. No additional post-modeling structural refinement of the models was performed. The models generated from each Lipid flipping mechanism of MFS protein combination were evaluated by the discrete optimized protein energy score (39), and the 20 best ranking models were selected (total 600 models). R.m.s. deviation values between each template and their corresponding models (top 20) were calculated and average values corresponding to each alignment method used are summarized in Table S1. We also included models predicted by HHpred (10 models), Phyre2 (20 models), Rap-torX (5 models), and I-TASSER (4 models), based on single templates. HHpred, Phyre2, and RaptorX also created one model each based on multiple templates that were also included for validation. In total, the 42 models generated by the four protein structure prediction programs along with the 600 models independently generated using MODELLER were considered for further validation.

Structural validation
Because of the very low sequence homology shared by the chosen templates and query, selection of the best structural model from among the 642 models for further interpretation solely based on parameters such as template query sequence identity, similarity, r.m.s. deviation, and model discrete optimized protein energy scores would be flawed and incomplete. To improve our confidence in the accuracy of a given model, we docked known substrates and nonsubstrates to the models and analyzed the docked scores and conformations. We reasoned that the more accurate a given model is the better it could be at distinguishing substrates from nonsubstrates. Thus, each of the 642 models was docked with a chemical library containing LPE and LPG, LPC, and LPA using DOCK v3.6 (19). The docked poses for each of the four compounds were scored and ranked on the basis of the docking energy function. The final model that was generated by MODELLER based on the template structure of a putative MFS transporter YajR Ec protein using the alignment created with PRALINE was chosen based on the best docking score qualitatively matching the substrate selectivity preferences (LPE Ϸ LPG Ͼ LPA Ͼ Ͼ LPC) previously reported for LplT Kp (8).