The Lipid A 1-Phosphatase, LpxE, Functionally Connects Multiple Layers of Bacterial Envelope Biogenesis

Dephosphorylation of the lipid A 1-phosphate by LpxE in Gram-negative bacteria plays important roles in antibiotic resistance, bacterial virulence, and modulation of the host immune system. Our results demonstrate that in addition to removing the 1-phosphate from lipid A, LpxEs also dephosphorylate undecaprenyl pyrophosphate, an important metabolite for the synthesis of the essential envelope components, peptidoglycan and O-antigen. Therefore, LpxEs participate in multiple layers of biogenesis of the Gram-negative bacterial envelope and increase antibiotic resistance. This discovery marks an important step toward understanding the regulation and biogenesis of the Gram-negative bacterial envelope.

at the outer leaflet of the outer membrane through the hydrophobic lipid A moiety. As peptidoglycan, phospholipids, and LPS are synthesized through distinct pathways, how Gram-negative bacteria orchestrate the biogenesis and remodeling across three layers of the cell envelope for optimal bacterial growth and virulence remains incompletely understood.
As the major lipid species coating the outer surface of Gram-negative bacteria, lipid A is the predominant signaling molecule that is detected by the mammalian Toll-like receptor 4 (TLR4)/myeloid differentiation factor 2 (MD-2) innate immune receptor (1) and caspase-4/-5/-11 (2) to trigger the host innate immune response to bacterial infection. With few exceptions, Gram-negative bacteria constitutively synthesize the 1,4=-bisphosphorylated tetra-acyl-lipid A intermediate, 3-deoxy-D-manno-oct-2ulosonic acid (Kdo)-linked lipid IV A (Kdo 2 -lipid IV A ), via the action of seven conserved enzymes in the Raetz pathway (3) (see Fig. S1A in the supplemental material), which are essential to nearly all Gram-negative bacteria and are attractive targets for novel antibiotics (4)(5)(6). Gram-negative bacteria additionally harbor modification enzymes that further process the Kdo 2 -lipid IV A intermediate to generate unique lipid A molecules in each bacterial species to adapt to environmental changes and evade the host immune response (7). For example, the lipid A 1-phosphate is a key determinant for lipid A recognition by the mammalian TLR4/MD-2 innate immune receptor (8). Removal of the lipid A 1-phosphate by the membrane-embedded phosphatase LpxE strongly protects bacteria against host cationic peptides and the last-resort antibiotic colistin (9), significantly dampens the host innate immune response, and dramatically increases colonization and survival of Helicobacter pylori in the gastric mucosa (10).
In order to gain molecular insights into the structure and function of the lipid A 1-phosphatase LpxE, we identified the previously uncharacterized gene aq_1706 from Aquifex aeolicus as the gene for the thermophilic LpxE enzyme (LpxE AA ). Our structural analysis of LpxE AA shows distinct features between LpxE AA and Escherichia coli PgpB (PgpB EC ) enzymes but reveals a surprising structural similarity to YodM, a phosphatase of phosphatidylglycerol phosphate (PGP) in the Gram-positive bacterium Bacillus subtilis with a weak in vitro activity on undecaprenyl pyrophosphate (C 55 -PP). Consistent with our structural analysis, we found that LpxE AA possesses substantial in vitro activities toward Kdo 2 -lipid A/lipid IV A , C 55 -PP, and PGP and complements E. coli strains deficient in C 55 -PP phosphatase and PGP phosphatase activities. In addition to the LpxE enzyme from A. aeolicus, distant LpxE orthologs from Francisella, Helicobacter, and Rhizobium also complement E. coli strains deficient in the C 55 -PP phosphatase activity, supporting the notion that the multifunctional lipid phosphatase activity is a general feature of LpxE enzymes. Significantly, deletion of the native lpxE gene sensitizes Francisella novicida to bacitracin, an antibiotic that sequesters C 55 -PP to disrupt peptidoglycan synthesis; furthermore, suppression of plasmid-encoded lpxE in the F. novicida strain deficient in the endogenous C 55 -PP phosphatase activity results in noticeable changes in cell morphology, profound reduction of O-antigen repeats in LPS, and loss of cell viability. Taken together, these observations reveal a previously unappreciated contribution of LpxE to peptidoglycan biogenesis and LPS O-antigen modification beyond its well-recognized role as the lipid A 1-phosphatase to orchestrate the remodeling of multiple layers of the Gram-negative bacterial envelope to respond to environmental changes, evade host immune surveillance, and promote bacterial viability and virulence.

RESULTS
A distant ortholog of LpxE FN in A. aeolicus. LpxE is a member of the lipid phosphatase/phosphotransferase (LPT) family, a well-distributed family of lipidprocessing enzymes also known as the integral transmembrane branch of the type II phosphatidic acid phosphatase (PAP2) superfamily (11,12). This family is characterized by a conserved tripartite active site motif of KX 6 RP---PSGH---SRX 5 HX 3 D and activity independent of Mg 2ϩ or other cations (13). The LPT family includes enzymes responsible for processing several types of lipids in Gram-negative bacteria, including the membrane-embedded PgpB, which dephosphorylates PGP and C 55 -PP (14) (Fig. S2). Even though PgpB and LpxE are both members of the LPT family, they have been reported to have distinct substrate specificities: PgpB is unable to utilize lipid A as a substrate (15), whereas purified LpxE from Rhizobium leguminosarum (LpxE RL ) utilizes PGP ϳ1,000 times less efficiently than lipid A species as a substrate in vitro (16).
In order to gain a molecular understanding of the LpxE structure and function, we searched for a thermophilic LpxE enzyme from Aquificae to facilitate structural analysis. The lipid A of Aquifex pyrophilus LPS contains D-galacturonic acid in place of phosphates at the 1-and 4=-positions (17) (Fig. S1B). As the 1,4=-bisphosphorylated lipid IV A is a common lipid A intermediate before further modification (18,19) and as Aquificae has the conserved biosynthetic enzymes to make 1,4=-bisphosphorylated lipid IV A , incorporation of the D-galacturonic acid moiety requires the removal of 1-phosphate from lipid A, indicating the presence of the lipid A 1-phosphatase activity in Aquificae. Such a rationale led us to search for the gene responsible for the lipid A 1-phosphatase activity in A. aeolicus VF5, as no lipid A 1-phosphatase has been reported in any Aquifex species.
A  16.58% and 14.57%, respectively), except for the well-conserved tripartite active-site motif of KX 6 RP---PSGH---SRX 5 HX 3 D (Fig. S3). In order to determine if aq_1706 encodes the lipid A 1-phosphatase activity in vivo, we overexpressed Aq_1706 in the heptose transferase-deficient E. coli strain WBB06, which produces Kdo 2 -lipid A instead of full-length LPS, to facilitate mass spectrometry analysis of lipid A modifications (21). Since E. coli does not encode LpxE activity, mass spectrometry analysis of the extracted lipids showed normal lipid A containing 1-phosphate with an m/z of 1,117.633 for the [M-2H] 2Ϫ ion species (calculated m/z, 1,117.661 for the exact mass of 2,237.336 of Kdo 2 -lipid A) from E. coli cells expressing a control vector; in contrast, overexpression of Aq_1706 in E. coli led to the disappearance of the intact lipid A species and significant accumulation of lipid A molecules lacking the 1-phosphate group, with an m/z of 1,077.647 for the [M-2H] 2Ϫ ion species (calculated m/z, 1,077.678 for the exact mass of 2,157.370 of 1-dephospho Kdo 2 -lipid A), consistent with the anticipated lipid A 1-phosphatase activity (Fig. 1A). In order to verify that the loss of phosphate occurred at the 1-position, but not at the 4=-position, we further tested the ability of LpxE to dephosphorylate 4=-32 P-labeled Kdo 2 -lipid IV A , which was previously shown to be an efficient substrate for LpxE enzymes with specific activity comparable to that for the substrate Kdo 2 -lipid A (16). We found that treatment of Kdo 2 -[4=-32 P] lipid IV A with membrane extracts from E. coli overexpressing Aq_1706, but not those carrying a control vector, resulted in time-dependent reduction of the Kdo 2 -lipid IV A band and accumulation of an upper-shifted band on the thin-layer chromatography (TLC) plate (Fig. 1B), reflecting the removal of 1-phosphate but retention of the 32 P-labeled 4=-phosphate group. Taken together, these observations verify aq_1706 in A. aeolicus as the gene that encodes the thermophilic lipid A 1-phosphatase LpxE (LpxE AA ).
Structural analysis of LpxE AA reveals a striking similarity to YodM in B. subtilis. After verifying the lipid A 1-phosphatase activity of LpxE AA , we cloned and purified LpxE AA . Consistent with the TMHMM analysis (http://www.cbs.dtu.dk/ services/TMHMM/), high-yield expression of LpxE AA was achieved in a maltose-binding protein (MBP) fusion construct containing an N-terminal PelB secretion signal (22), suggesting that the N terminus of LpxE AA is located at the periplasmic side of the inner membrane. The crystal structure of LpxE AA containing an I63M mutation was determined at 2.38 Å ( Fig. 2A; statistics shown in Table S1). The selenomethionine substitution of the nonconserved I63 residue (I63M) was designed to enhance the selenium single anomalous dispersion (Se-SAD) signal for de novo phasing. The overall structure of LpxE AA contains seven ␣-helices, including an N-terminal amphiphilic helix lying at the periplasmic surface of the inner membrane and five tightly packed transmembrane helices (␣3 to ␣7). Apart from ␣2, which originates from the periplasmic surface and penetrates halfway across the inner membrane at an ϳ45°angle and immediately connects to transmembrane helix ␣3, the remaining helices are oriented largely in parallel or antiparallel with each other and perpendicularly to the membrane plane. Looking from the periplasmic surface, helix 5 (␣5) is located at the center, which is surrounded by ␣2, ␣3, ␣4, ␣7, and ␣6 in a counterclockwise fashion (Fig. 2B).
The active site of LpxE is located at the periplasmic surface of the inner membrane and is defined by conserved motifs specific to the PAP2 enzymes (K 73 X 6 R 80 P---R 137 X 5 H 143 X 3 D 147 ) located at the C-terminal end of ␣4, the ␣4-␣5 loop, ␣6, the ␣6-␣7 loop, and the N terminus of ␣7 (Fig. 2C). Fortuitously, a sulfate molecule is found in the active site, which is a structural analog of the 1-phosphate group of lipid A. The sulfate group is extensively recognized by K73 and R80 of the K 73 X 6 R 80 P motif and R137 of the R 137 X 5 H 143 X 3 D 147 motif. The catalytically important H143 is located 3.3 Å away from the sulfur atom of the sulfate group, ready to carry out inline attack to remove the phosphate group of the lipid substrate. D147, the last residue of the R 137 X 5 H 143 X 3 D 147 motif, forms a hydrogen bond with H143. Although the corresponding aspartate residue is found in most LpxE enzymes (Fig. S3), it is absent in the LpxE ortholog from H. pylori (LpxE HP ), suggesting that it is not absolutely required for catalysis. The first three residues of the PSGH motif are conserved in LpxE AA , with the central serine residue (S101) serving as a helix cap to stabilize helix ␣5, but the histidine residue is replaced with an aspartate residue in LpxE AA (Fig. 2C).
The LpxE AA structure shows noticeable conformational discrepancy with the previously reported structures of PgpB EC (PDB codes 4PX7 and 5JWY) (23,24), another PAP2 family enzyme, with overall backbone root mean square deviations (RMSDs) of ϳ4.5 Å (Fig. S4); surprisingly, LpxE AA is structurally similar to the recently reported YodM in B. subtilis (PDB code 5JKI) (25), a PGP phosphatase with a weak in vitro activity on C 55 -PP, with an overall backbone RMSD of 1.2 Å (Fig. 2D). The major differences of these two enzymes are the absence of an N-terminal transmembrane helix in LpxE AA in comparison with YodM, a longer ␣4-␣5 loop in LpxE AA , and a significant conformational variation of the ␣4-␣5 loop surrounding the active site.
LpxE AA is a trifunctional lipid phosphatase in vitro and functionally complements E. coli mutants deficient in C 55 -PP or PGP phosphatase activities. Surprised The sulfate molecule is shown in the stick model. Side chains of H143 and D147 from the RX 5 HX 3 D motif and conserved residues coordinating the sulfate molecule, including K73 and R80 from the KX 6 RP motif and R137 of the RX 5 HX 3 D motif, are shown in the stick model. Hydrogen bonds are shown by dashed lines. The sulfate group is additionally stabilized by the interaction with the electrical dipole of helix ␣5 (indicated by gray hydrogen bonds). The conserved PSG motif is colored in coral, with C␣ atoms shown in spheres. (D) Superimposition of LpxE AA (cyan) with YodM BS (PDB code 5JKI; pink), revealing striking structural similarities. The major differences between the two structures are highlighted, with dashed circles indicating missing structural features (helix or loop) and arrows indicating conformational discrepancy.
LpxE Plays Multiple Roles in Cell Envelope Biogenesis ® by the structural similarity between LpxE AA and YodM BS , we asked whether LpxE AA could function as a C 55 -PP and PGP phosphatase. To address this question, we compared the specific activities of purified LpxE AA toward Kdo 2 -lipid A, PGP, and C 55 -PP using the malachite green assay to detect the release of inorganic phosphate. As expected, LpxE AA efficiently catalyzed the hydrolysis of 1-phosphate from Kdo 2 -lipid A, with a specific activity of 2.04 Ϯ 0.46 mol/mg/min. Moreover, LpxE AA catalyzed C 55 -PP more efficiently than it catalyzed Kdo 2 -lipid A, with a specific activity of 3.58 Ϯ 0.47 mol/mg/min-a value that is ϳ1.8-fold higher than that toward Kdo 2 -lipid A. Finally, LpxE AA also displayed significant activity toward PGP, with a specific activity of 0.75 Ϯ 0.11 mol/mg/min, ϳ40% of its activity toward Kdo 2 -lipid A ( Table 1). Taken together, our biochemical assays validate LpxE AA as a trifunctional LPT enzyme that efficiently dephosphorylates chemically diverse Kdo 2 -lipid A (glycolipids), PGP (phosphoglycerol lipid), and C 55 -PP (isoprenyl lipid) in vitro.
In order to obtain further evidence of the trifunctional role of LpxE AA in cells, we examined whether LpxE AA could functionally rescue lethal E. coli mutants lacking C 55 -PP phosphatase or PGP phosphatase activities. E. coli contains four C 55 -PP phosphatases, BacA, PgpB, YbjG, and LpxT. A deletion mutant, ΔybjG ΔbacA ΔpgpB::kan, in E. coli is lethal unless rescued by a plasmid expressing BacA, PgpB, or YbjG (26). To examine if LpxE AA could function as a C 55 -PP phosphatase in cells, we set up complementation of the lethal ΔybjG ΔpgpB ΔbacA::kan E. coli mutant carrying lpxE AA on a low-copy-number, temperature-sensitive pMAK705 vector (pMAK-lpxE AA ). The E. coli bacA gene, encoding the C 55 -PP phosphatase, was used as the positive control (pMAK-bacA EC ). We found that overexpression of LpxE AA and BacA EC from pMAK705-derived plasmids complemented the lethal phenotype of the ΔybjG ΔpgpB ΔbacA::kan triple knockout in E. coli on an LB agar plate at 30°C; such a complementation effect was lost when cells were grown at 42°C, consistent with the loss of the temperature-sensitive pMAK705 plasmid encoding LpxE AA or BacA EC and confirming that LpxE AA functionally complements the loss of C 55 -PP phosphatase activity in E. coli (Fig. 3A).
We similarly tested whether LpxE AA functionally complements the loss of PGP phosphatase activity in E. coli. E. coli has three PGP phosphatases, PgpA, PgpB, and PgpC (27). A ΔpgpA ΔpgpB ΔpgpC::kan triple-knockout mutant is lethal unless it is rescued by a plasmid harboring an active PGP phosphatase (27). Overexpression of LpxE AA or the positive control PgpA EC from the temperature-sensitive pMAK705 plasmid supported the growth of the ΔpgpA ΔpgpB ΔpgpC::kan triple-knockout mutant strains at 30°C but not at 42°C. In contrast, the control strain (W3110/pMAK705) grew well at both temperatures (Fig. 3B). These observations confirm that LpxE AA is a functional PGP phosphatase in E. coli.
While the pMAK705 vector-encoded LpxE AA complemented E. coli triple knockouts lacking C 55 -PP phosphatase or PGP phosphatase activities, pMAK705 has a higher copy number (pSC101 origin, ϳ5 copies/cell) than that of the chromosome in E. coli (single copy/cell). In order to mitigate the concern that the observed genetic complementation was caused by multiple copies of the lpxE AA gene, we replaced the pgpB gene in the chromosome of E. coli (BW25113) ΔybjG ΔbacA with a gene cassette (P L -lpxE AA -FRTkan-FRT) containing lpxE AA and a kanamycin resistance gene under the control of the P L promoter (28). The resulting E. coli strain (E. coli BW25113 ΔybjG ΔbacA ΔpgpB::P L -lpxE AA -FRT-kan-FRT) grew on an LB agar plate, and the proper knockouts of bacA, ybjG, and pgpB were verified by PCR (Fig. 3C), confirming that the chromosomal copy of lpxE AA complemented the loss of C 55 -PP phosphatase activity. Using a similar approach,  we also replaced the pgpB gene of E. coli (W3110) ΔpgpA with P L -lpxE AA -FRT-kan-FRT, removed the kanamycin resistance cassette (29), and then knocked out pgpC. The resulting strain (E. coli W3110 ΔpgpA ΔpgpB::P L -lpxE AA ΔpgpC::kan) also grew on an LB agar plate, and knockouts of pgpA, pgpB, and pgpC were verified by PCR (Fig. 3D), confirming that the chromosomal copy of lpxE AA similarly complemented the loss of PGP phosphatase activity. Altogether, the substantial phosphatase activities of LpxE AA toward Kdo 2 -lipid A, C 55 -PP, and PGP in vitro and its ability to complement the loss of C 55 -PP and PGP phosphatase activities in E. coli-both via the plasmid-borne gene and via chromosomal knock-in-strongly support the multifunctionality of LpxE AA in Gram-negative bacterial envelope biogenesis.
LpxE FN is a bifunctional lipid phosphatase in vitro and functionally complements an E. coli mutant deficient in the C 55 -PP phosphatase activity. Despite the intriguing observation of the multifunctionality of LpxE AA , it is challenging to establish the biological consequence in its native host due to the difficulty of culturing and genetic manipulation of A. aeolicus. Therefore, we asked if other LpxE enzymes from genetically trackable bacteria similarly display multifunctional lipid phosphatase activities. In order to answer this question, we chose LpxE FN , a distant ortholog of LpxE AA , for further characterization. The ability of LpxE FN to dephosphorylate lipid A at the 1-position was previously reported (15), but its activity toward other lipid substrates has not been thoroughly investigated. We first conducted similar complementation experiments using E. coli strains deficient in either the C 55 -PP phosphatase activity or PGP phosphatase activity carrying the temperature-sensitive pMAK-lpxE FN . We found that LpxE FN complemented the loss of C 55 -PP phosphatase activity of E. coli (ΔybjG ΔpgpB ΔbacA::kan) at 30°C but not at 42°C, indicating that LpxE FN is a functional C 55 -PP phosphatase in E. coli (Fig. 4A). However, we were unable to complement E. coli deficient in the PGP activity (ΔpgpA ΔpgpB ΔpgpC::kan) with a plasmid encoding LpxE FN (pMAK-lpxE FN ). Consistently, we found that purified LpxE FN displayed significant phosphatase activity toward both Kdo 2 -lipid A and C 55 -PP and processed these two substrates with similar efficiencies (specific activities of 3.25 Ϯ 0.21 mol/mg/min for Kdo 2 -lipid A and 2.99 Ϯ 0.45 mol/mg/min for C 55 -PP), but its activity toward PGP was ϳ100-fold lower (specific activity of 0.038 Ϯ 0.009 mol/mg/min) ( Table 1), confirming that LpxE FN is a bifunctional lipid phosphatase.
F. novicida harbors two C 55 -PP phosphatases: LpxE FN and FTN_1552. It is important to note that the lipid A 1-phosphatase activity is not essential in bacteria but the C 55 -PP phosphatase activity is. Prior to this study, no enzyme encoding the C 55 -PP phosphatase activity had been identified in F. novicida. As the transposon mutant of lpxE FN is not lethal in F. novicida (30), we reasoned that there must exist another enzyme encoding the C 55 -PP phosphatase activity in F. novicida. By searching for F. novicida proteins homologous to E. coli enzymes containing C 55 -PP phosphatase activity (i.e., BacA EC , YbjG EC , PgpB EC , and LpxT EC ) using PSI-BLAST (20), we have identified a PAP2 family protein of unknown function, FTN_1552, as a potential candidate of the C 55 -PP phosphatase (PSI-BLAST of PgpB EC : E value of 0.003 and sequence identity of 16.47%). We found that the temperature-sensitive pMAK705 vector harboring ftn_1552 complemented the E. coli strain deficient in C 55 -PP phosphatase activity (ΔybjG ΔpgpB ΔbacA::kan), confirming ftn_1552 as the gene encoding the C 55 -PP phosphatase activity (Fig. 4A). FTN_1552 was subsequently renamed UppP FN . Purified UppP FN appears to be a specific enzyme for C 55 -PP, with a specific activity of 22.71 Ϯ 2.62 mol/ mg/min, and displays little activity toward Kdo 2 -lipid A and PGP (specific activities of 0.010 Ϯ 0.005 mol/mg/min and 0.031 Ϯ 0.007 mol/mg/min, respectively [ Table 1]). Importantly, while F. novicida strains containing a chromosomal deletion of either lpxE FN or uppP FN were viable, we were unable to generate F. novicida strains containing both deletions (ΔlpxE FN ΔuppP FN ) in the chromosome (Fig. 4B and C). However, F. novicida cells were viable in the ΔuppP FN background when lpxE FN was replaced with lpxE AA (Fig. 4B and C). Furthermore, when F. novicida was first transformed with a plasmid (pEDL17) bearing lpxE FN under the control of an anhydrotetracycline (aTc) promoter, we were also able to obtain viable F. novicida colonies containing chromosomal deletions of both lpxE FN and uppP FN (U112 ΔlpxE FN ΔuppP FN ::kan/pEDL17-lpxE FN ). The presence of proper chromosomal deletions of the lpxE FN and uppP FN genes was verified by PCR (Fig. 4D). As expected, the viability of such a strain depends on the expression of plasmid-encoded LpxE FN : prolonged withdrawal of aTc suppressed the bacterial growth and slowly resulted in cell lysis in culture (Fig. 4D), reinforcing the notion that UppP FN and LpxE FN share redundant C 55 -PP phosphatase activities in Francisella.
LpxE FN functionally connects multiple layers of envelope biogenesis in F. novicida. After establishing that LpxE FN shares C 55 -PP phosphatase activity with UppP FN , we further examined the biological implication of the multifunctional enzymatic activity of LpxE FN in its native host, F. novicida. We first verified the role of LpxE FN LpxE Plays Multiple Roles in Cell Envelope Biogenesis ® as a lipid A 1-phosphatase. Wild-type (WT) F. novicida cells contain both LPS (i.e., core oligosaccharide and O-antigen-modified Kdo-lipid A3 without 1-and 4=-phosphates) and free lipid A species A1 and A2, which do not contain core oligosaccharides/Kdo or O-antigen (lipid A2 differs from lipid A1 in that it has an additional ␣-linked glucose moiety attached to its 6=-position; also see the schematic lipid A structures of WT F. novicida in Fig. 5A) (31). Both lipid A1 and lipid A2 are further modified by FlmK, which transfers galactosamine from C 55 -P-galactosamine to the 1-phosphate of lipid A (32,33). As the core oligosaccharide and O-antigen-modified lipid A are inefficiently extracted by the Bligh-Dyer method for mass spectrometry analysis, we examined the effect of ΔlpxE FN in the F. novicida strain deficient in the glycosyltransferase activity (ΔlpcC), which produces Kdo-lipid A3 (instead of LPS), in addition to lipids A1 and A2 found in the wild-type cells (Fig. 5A). Accumulations of Kdo-(1-phospho)-lipid A3 and Kdo-(galactosamine-1-phospho)-lipid A3, as well as the disappearance of Kdo-lipid A3, were observed in F. novicida when lpxE FN was deleted (ΔlpxE FN ), confirming the lipid A 1-phosphatase activity of LpxE FN in cells (Fig. 5A and Fig. S5).
As the C 55 -PP phosphatase activity of LpxE FN (2.99 Ϯ 0.45 mol/mg/min) is only ϳ7-fold smaller than that of UppP FN (22.71 Ϯ 2.62 mol/mg/min), we asked whether LpxE FN could functionally contribute to the bacterial envelope biogenesis beyond lipid A modification at the 1-phosphate position. We first compared the sensitivities of the wild-type F. novicida strain (U112) and mutant strains containing either the lpxE or uppP deletion to bacitracin, an antibiotic sequestering C 55 -PP. We found that while the loss of uppP in F. novicida generated an 8.5-fold drop of MIC in comparison with that of the WT strain (0.5 M versus 4.25 M), as expected, the loss of lpxE also resulted in ϳ1.7-fold drop of the MIC of bacitracin (2.5 M for F. novicida ΔlpxE) (Fig. 5B), implicating a functional role of LpxE FN in the recycling of C 55 -PP.
In order to isolate the biological effect of LpxE FN , we utilized the F. novicida strain containing chromosomal deletions of both uppP FN and lpxE FN , which is complemented by a plasmid carrying lpxE FN under the control of an aTc promoter. We found that the loss of plasmid-mediated expression of LpxE FN due to withdrawal of aTc in the growth medium resulted in cell enlargement, reflecting defective peptidoglycan biosynthesis (Fig. 5C). Strikingly, while no change of O-antigen repeats was observed in F. novicida cells containing the chromosomal deletion of either lpxE or uppP in comparison with WT cells (Fig. 5D), transient suppression of LpxE led to a dramatic reduction of the LPS O-antigen repeats, including both high-and low-repeat species (Fig. 5E) (34), suggesting a contribution of LpxE to the O-antigen biogenesis. These observations are consistent with the notion that the biosynthesis and transport of peptidoglycan and O-antigen depend on C 55 -P, the product of LpxE FN (and UppP FN ) activity, and reveal a previously unappreciated function of LpxE in the biogenesis and remodeling of multiple components across the bacterial envelope: peptidoglycan, free lipid A, and the O-antigen repeat of LPS.

DISCUSSION
LpxE enzymes are important virulence factors that promote bacterial survival, fitness, and pathogenicity. In H. pylori and Rhizobium etli CE3, the chromosomal knockout of lpxE resulted in increased susceptibility to positively charged antimicrobial peptides such as polymyxin B and colistin (9,35), presumably due to the retention of 1-phosphate of lipid A. Previous studies showed that Rhizobium LpxE displays over a 1,000-fold preference of Kdo 2 -lipid A/lipid IV A over PGP; therefore, LpxE has been regarded as a highly specific monofunctional enzyme whose sole activity is to remove the 1-phosphate from lipid A. In this study, based on the striking structural similarity between LpxE AA and YodM BS , a PGP phosphatase with a weak in vitro activity on C 55 -PP phosphatase, we discovered that LpxE is a multifunctional lipid phosphatase. The LpxE enzyme from A. aeolicus displays significant activities toward Kdo 2 -lipid A/lipid IV A , C 55 -PP and PGP and functionally complements E. coli strains deficient in C 55 -PP or PGP phosphatase activities. Likewise, the LpxE enzyme from F. novicida is a dual-function enzyme that processes Kdo 2 -lipid A and C 55 -PP with similar efficiencies. Strikingly, deletion of lpxE FN in its native host F. novicida resulted in accumulation of phosphorylated lipid A species and increased sensitivity to bacitracin; in the C 55 -PP phosphatasedeficient (ΔuppP and ΔlpxE double-knockout mutant) F. novicida, suppression of LpxE FN expression from the plasmid resulted in cell deformation due to defective peptidogly-can biosynthesis and the loss of O-antigen repeats in LPS associated with reduced O-antigen transport, both of which are critically dependent on the recycling of C 55 -PP to C 55 -P. Taken together, these results show that LpxE enzymes from A. aeolicus and F. novicida functionally connect multiple layers of bacterial envelope biogenesis and remodeling. Such multiple functional roles are not unique to LpxE enzymes from A. aeolicus and F. novicida: we found that LpxE enzymes from H. pylori and R. leguminosarum also complemented E. coli deficient in C 55 -PP phosphatase activities (Fig. S6), suggesting that these LpxE enzymes can similarly process Kdo 2 -lipid A and C 55 -PP to synchronize lipid A modification with peptidoglycan biosynthesis and O-antigen modification of LPS.
It is appropriate to ask why LpxE has evolved into a multifunctional enzyme. There are several potential explanations. First, it is conceivable that the peptidoglycan biosynthesis is such an essential process that multiple enzymes, including LpxE, are employed as the backup enzymes for the C 55 -PP phosphatase-mediated recycling reaction for peptidoglycan charging and biosynthesis. Second, it is possible that LpxE from Aquifex species represents an ancestral lipid phosphatase, which, although primitive, is sufficient to conduct all lipid phosphatase activities to support the bacterial envelope biogenesis and remodeling, while other LPT family of lipid phosphatases, such as the PGP phosphatase, evolved later as specialized, highly efficient enzymes. Third, it is also likely that LpxE evolved as a multifunctional enzyme to coordinate lipid A modification and the biogenesis of other layers of bacterial envelope. As 1,4=bisphosphorylated lipid A chelates metal ions to form a fortified layer for bacterial protection, removal of the 1-phosphate could weaken the lipid A layer and increase membrane permeability. It is conceivable that the weakened lipid A layer is compensated by the elevated peptidoglycan biosynthesis and enhanced O-antigen decoration of LPS. Thus, bestowing LpxE with the multifunctionality toward Kdo 2 -lipid A and C 55 -PP (and, in the case of A. aeolicus, PGP) enables LpxE to orchestrate lipid A modification with bacterial envelope remodeling at multiple layers (Fig. 6) in order to promote the optimal bacterial growth and enhance bacterial survival in nature and the human host.
The Gram-negative bacterial envelope contains three layers. How Gram-negative bacteria coordinate the biogenesis and remodeling of different layers of the bacterial envelope has remained an area of active investigation. Our study has revealed the first biological evidence of a multifunctional enzyme, LpxE in F. novicida, that natively couples lipid A 1-dephosphorylation with C 55 -PP recycling to enhance peptidoglycan biogenesis and O-antigen decoration of LPS, promote cell viability against antimicrobial peptides, evade host immune surveillance, and ultimately support bacterial pathogenesis. We suggest that such a multifunctional role represents a common but previously unappreciated mechanism for Gram-negative bacteria to coordinate bacterial envelop biogenesis across different layers.

MATERIALS AND METHODS
Data collection and refinement statistics of LpxE AA are listed in Table S1. All strains and plasmids used in this work are listed in Tables S2 and S3, respectively.
Plasmid and strain constructions and growth conditions are described in the supplemental material in detail.
Extraction of lipid A species, TLC and mass spectrometry analyses of lipid A species, and assay conditions are described in the Supplementary Methods section of the supplemental material.
Characterizations of F. novicida U112 mutants are described in the Supplementary Methods section of the supplemental material.