Rickettsia Lipid A Biosynthesis Utilizes the Late Acyltransferase LpxJ for Secondary Fatty Acid Addition

Lipopolysaccharide (LPS) triggers an inflammatory response through the TLR4-MD2 receptor complex and inflammatory caspases, a process mediated by the lipid A moiety of LPS. Species of Rickettsia directly engage both extracellular and intracellular immunosurveillance, yet little is known about rickettsial lipid A. Here, we demonstrate that the alternative lipid A acyltransferase, LpxJ, from Rickettsia typhi and R. rickettsii catalyzes the addition of C16 fatty acid chains into the lipid A 3′-linked primary acyl chain, accounting for major structural differences relative to the highly inflammatory lipid A of Escherichia coli.

Gram-negative bacterial cell envelope (5)(6)(7). The inner membrane (IM) and outer membrane (OM) are separated by a relatively thin peptidoglycan layer (8), which contains diaminopimelate in the stem peptide (9). The OM is asymmetric, with the inner leaflet composed mainly of glycerophospholipids (10,11) and the outer leaflet predominantly composed of lipopolysaccharide (LPS) (12)(13)(14). As the interface connecting host to microbe, the rickettsial OM constituents are critically important not only for host cell invasion but also for mediating intracellular survival. The composition of surface proteins has been identified for several species (15)(16)(17)(18), and nearly two dozen effectors are recognized as components of the secretome (19); however, the nonproteinaceous surface components of Rickettsia, including LPS, are critically understudied.
LPS is highly antigenic in rickettsial infection (20,21) and is composed of an outer O-antigen polysaccharide linked to a core oligosaccharide, which is anchored in the bacterial outer leaflet by lipid A (22). Lipid A is a pathogen-associated molecular pattern (PAMP) known for its ability to trigger an inflammatory response through its interaction with the Toll-like receptor 4/myeloid differentiation factor 2 (TLR4/MD2) complex, as well as activation of the noncanonical inflammasome through cytosolic caspases (Casp-4/Casp-5 in humans; Casp-11 in mice) (23)(24)(25). In a mouse model of rickettsiosis, TLR4/MD2 activation is critical for bacterial clearance (26,27); however, little is known about the contribution of LPS to the inflammatory nature of Rickettsia infection (28,29). Recent analysis of the lipid A from Rickettsia typhi revealed structural differences relative to the highly inflammatory lipid A of Escherichia coli though the potential of rickettsial lipid A to act as a TLR4 agonist remains unclear (30) (Fig. 1).
Our recent phylogenomics study indicated that rickettsiae contain a nearly conserved Raetz pathway for the synthesis of lipid A, lacking only the late acyltransferase LpxM (4). LpxM, which catalyzes the 3= secondary acylation of 3-deoxy-D-manno-2octulosonic acid (KDO 2 )-lauroyl-lipid IV A (typically transferring myristate) subsequent to 2= secondary acylation (carried out by LpxL), is absent in many bacteria, some of which alternatively carry a nonorthologous late acyltransferase named LpxJ (previously named DUF374) (31)(32)(33). LpxJ enzymes of Helicobacter pylori, Campylobacter jejuni, and Wolinella succinogenes (all Epsilonproteobacteria pathogens) catalyze 3= secondary acylation but can or must do so prior to 2= secondary acylation (LpxL) and even 3-deoxy-D-manno-octulosonic acid (KDO) transfer (carried out by the KDO transferase WaaA). Thus, LpxJ enzymes can be considered more functionally promiscuous than their LpxM counterparts. As LpxJ homologs are present in all Rickettsia genomes, with the R. typhi enzyme sharing 27% identity with H. pylori LpxJ, we reasoned that these enzymes complete the Raetz pathway for rickettsial lipid A biosynthesis and incorporate a C 16 fatty acid chain as a 3= secondary acylation (30).
Here, we provide enzymatic evidence that Rickettsia LpxJ complements E. coli LpxM mutants and carries out 3= secondary acylation of lipid IV A and lauroyl-lipid IV A . Additionally, targeted mutagenesis based on comparative analysis of Ͼ2,800 DUF374 family members with LpxJ homologs reveals residues critical for acylation. In line with prior work (31), our data demonstrate that divergent LpxJ and LpxM active sites both catalyze 3= secondary acylation for lipid A biosynthesis and that LpxJ is a nonorthologous replacement of LpxM in a vast range of diverse bacteria. As lipid A architecture is fundamental to OM integrity in Gram-negative bacteria, our findings indicate that LpxJ may be important in maintaining ideal membrane dynamics to facilitate molecular interactions at the host-pathogen interface.

RESULTS
Rickettsia encodes a homolog of LpxJ. Rickettsial comparative genomic analysis has identified a nearly complete Raetz pathway of lipid A biosynthesis (see Fig. S1 in the supplemental material). However, Rickettsia species do not encode any enzymes similar to LpxM (also known as MsbB). Since R. typhi (and probably all species of Rickettsia) produces hexa-acylated lipid A (30), we reasoned that a lipid A acyltransferase analogous to LpxM has escaped gene annotation within rickettsial genomes. In a report describing LpxJ, Rubin et al. identified a putative LpxJ homolog in R. rickettsii (27% similarity at the protein level) (31). We have further identified LpxJ family genes throughout the genus Rickettsia (Table 1) and have selected putative homologs from R. typhi (RT0047) (Fig. S2) and R. rickettsii (A1G_00705), here termed LpxJ Rt and LpxJ Rr , respectively, for molecular characterization.  LpxJ Rt and LpxJ Rr complement an E. coli LpxM mutant. In order to investigate the role of LpxJ in Rickettsia lipid A biosynthesis, we utilized a heterologous system in which acylation-deficient lipid A mutants of E. coli act as a reporter of enzyme function for exogenously expressed acyltransferases. We first expressed LpxJ Rt and LpxJ Rr in the lpxM mutant MLK1067 that elaborates predominately penta-acylated lipid A. After expression of rickettsial proteins was induced (Fig. S3), lipid A extractions were prepared and subjected to matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) analysis to determine if rickettsial LpxJ can complement the loss of lpxM and produce hexa-acyl lipid A. In comparison to results in untransformed MLK1067, we observed additional lipid A species of increased mass from cells expressing LpxJ Rt and LpxJ Rr but no change from cultures transformed with an empty plasmid vector (Fig. 2). The ions at m/z 1,797 and m/z 1,825 represent the addition of C 14 (Δm/z 210) or C 16 (Δm/z 238), respectively, to the parental penta-acylated lipid A (m/z 1,587). MALDI-TOF MS results for LpxJ Rt were confirmed using gas chromatography (GC). Fatty acid peaks were identified by comparison to commercially available bacterial acid methyl ester (BAME) standards. The amount of each fatty acid present in lipid A was calculated by comparison to an internal pentadecanoic acid (C 15 ) standard over three biological replicates. LpxJ Rt -overexpressing cells show a 4-fold increase in the total amount of C 14 and C 16 (Fig. 3). Taken together, these data indicate that LpxJ Rt and LpxJ Rr are bona fide acyltransferases in the lipid A pathway.
LpxJ Rt activity is independent of LpxL activity. In E. coli lipid A synthesis, the final fatty acid addition by LpxM requires the prior activity of LpxL, the 2= secondary acyltransferase (KDO 2 -lauroyl-lipid IV A substrate) (34). In contrast to LpxM's rigid substrate selection, LpxJ has flexibility in its activity. LpxJ of H. pylori can act independently of LpxL activity, while homologs from C. jejuni and W. succinogenes act exclusively on tetra-acylated substrates. LpxJ Rt is shown above to act upon penta-acylated lipid A molecules ( Fig. 2), but it is unclear whether this is the only lipid A precursor that is a substrate. To determine the requirement of LpxL activity on the activity of LpxJ, we expressed LpxJ in E. coli strain MKV15b (35), which produces mostly tetra-acylated lipid A lacking secondary acylation on the 2= and 3= fatty acids (Fig. 4). We found that LpxJ Rt acts upon tetra-acylated substrate (m/z 1,404), transferring C 14 or C 16 fatty acids (m/z 1,615 and 1,642, respectively). These data suggest that Rickettsia lipid A synthesis does not share the strict operational order of acyl chain incorporation found in E. coli.
Global analysis reveals conserved LpxJ residues critical for acyl transfer. Analysis of 2,842 distinct LpxJ homologs identified four conserved regions (Fig. 5A). Interestingly, none of these regions contained an HX 4 D/E motif, the hallmark acid/base catalytic mechanism, or charge relay system, that defines glycerol-3-phosphate acyltransferase (GPAT), lysophosphatidic acid acyltransferase (LPAAT), dihydroxyacetonephosphate acyltransferase (DHAPAT), and 2-acyl-glycerophosphoethanolamine acyltransferase (LPEAT) (36). To gain insight on residues possibly comprising an alternative charge relay system in the LpxJ Rt active site, we mutated conserved His [H61A and H84(A/S)] and Asp (D86A, D132A, and D173A) residues within these regions, as well as two highly conserved Trp residues (W60A and W172A). Aside from H84(A/S), these mutants lacked the enzymatic function of LpxJ Rt when expressed in MLK1067, reverting the lipid A phenotype to that of the background strain (penta-acyl, m/z 1,587) (Fig. 5B). The negligible effect on LpxJ Rt activity observed by mutating H84 to either Ala or Ser (data not shown) indicates that H61 is the likely catalytic base of LpxJ.   Table S1 in the supplemental material.    The histidine, aspartic acid, or tryptophan residue at the indicated position in the primary sequence of LpxJ was mutated to alanine, and each construct was individually expressed in MLK1067. Loss of these highly conserved residues abolished the acyltransferase activity of the enzyme, reverting the lipid A phenotype to that of the background strain (penta-acyl, m/z 1,587). Mutation of histidine at position 84 to alanine or serine had no effect on enzymatic activity (data not shown). (C) Comparison of R. typhi LpxJ to four divergent lipid acyltransferases. Proteins with associated structures were obtained from the Protein Data Bank: Mycobacterium smegmatis PatA (PDB code 5F34), Acinetobacter baumannii LpxM (5KNK), Cucurbita moschata PlsB (1IUQ), and Thermotoga maritima PlsC (5KYM). R. typhi LpxJ was modeled to all four acyltransferase structures using Phyre2 (58) and fitted to an existing structural alignment template (37), which follows the convention established for naming conserved blocks within GPAT, LPAAT, DHAPAT, and LPEAT acyltransferases (59,60). Active-site residues for each structure are colored. For LpxJ, Asp132 is proposed to participate in the active site charge relay system with His61 (36,61).

Analysis of Rickettsia Lipid A Acyltransferase LpxJ Journal of Bacteriology
Provided that all three Asp mutants abolished LpxJ activity, we compared LpxJ to four divergent lipid acyltransferases for which structures have been solved (Fig. 5C). Modeling LpxJ Rt (RT0047) to these structures consistently positioned H61 within the canonical charge relay system (block I). Fitting LpxJ Rt to a structural alignment template (37) positioned D132 with other negatively charged residues in PatA (block III). This suggests that LpxJ is similar to lipid acyltransferases, which have diverged from the canonical HX 4 (D/E) motif by completing the charge relay system across blocks I and III (Fig. 5C). Unlike PatA, however, LpxJ does not have a second positively charged residue (block II) coordinating in the active site, which likely explains the lack of a conserved Asp or Glu within block I for LpxJ. This indicates further divergence of its charge relay system and distinguishes LpxJ from any known lipid acyltransferase.

DISCUSSION
Rickettsia species are obligate intracellular bacterial parasites that produce a typical Gram-negative envelope with the IM and OM separated by a thin peptidoglycan layer (4,5). Considering the invasive lifestyle of these parasites, the OM sits at the nexus of the host-pathogen interface and is rich in LPS, the classical Gram-negative PAMP (24). Understanding the interplay between mammalian and bacterial molecules at the host-pathogen interface is critically important for the development of novel therapeutic approaches. Considering the historical role of LPS as a mediator of inflammation and its location at the front line of host-pathogen interactions, it is safe to assume that LPS plays a role during Rickettsia infection. However, despite the apparent importance of LPS, there exists a paucity of information about the biology of LPS biosynthesis and its contribution to virulence. In order to address this gap in our understanding of this molecule, we have begun foundational work elucidating the basics of LPS biogenesis in Rickettsia. Here, we characterize new members of the recently discovered LpxJ family of lipid A acyltransferases from R. typhi and R. rickettsii. Using an E. coli reporter system, we have identified LpxJ Rt and LpxJ Rr to be lipid A acyltransferases in Rickettsia. These enzymes catalyze the transfer of secondary fatty acids, predominately C 14 or C 16 , to the 3-hydroxyl group of the 3= primary acyl chain. Interestingly, LpxJ Rt shows no preference for either penta-or tetra-acylated lipid A in our reporter system. This implies that activity of the enzyme is bispecific for either lipid IV A or acyl-lipid IV A and can act before LpxL (RT0704), similar to the previously characterized LpxJ homolog from H. pylori. These data indicate that multiple paths exist for secondary acylation of Rickettsia lipid A (Fig. 6) and that secondary fatty acid diversity can be more extensive than described by Fodorová et al. (30). Our analysis of R. typhi lipid A shown in Fig. S4 in the supplemental material indicates a minor lipid A peak at m/z 1,907 that is characteristic of shorter fatty acid chain incorporation than that of the major peak at m/z 1,936 depicted in Fig. 1. It is conceivable that the C 14 /C 16 promiscuity observed for LpxJ activity can be contributing to this in vivo heterogeneity. Further, both LpxJ homologs characterized here catalyze the addition of fatty acids with odd chain lengths, represented by the peak at m/z 1,810 in Fig. 2A and B. This addition is likely an artifact as neither E. coli nor Rickettsia spp. naturally produce odd chain fatty acids (4,38). It is possible that the presence of short-chain fatty acid metabolites in the growth medium causes nonnatural fatty acid synthesis leading up to cell lysis, as E. coli does produce some odd-chain-length fatty acids when grown in the presence of propionic acid (39). It is also possible that overexpression of rickettsial LpxJ in a heterologous E. coli system is influencing fatty acid biosynthesis and causing erroneous incorporation of oddlength, short-chain fatty acids as building blocks instead of the typical condensation of malonyl-coenzyme A (CoA) during fatty acid manufacturing.
The requirement for Kdosylation of lipid IV A prior to LpxJ Rt activity was not tested here. The E. coli pathway requires KDO sugars be present on lipid IV A prior to secondary acylation by LpxL and LpxM (40). However, previously characterized LpxJ homologs did not share this requirement (31). The combination of fatty acid selection (C 14 -C 16 ) and substrate promiscuity (tetra/penta-acyl lipid IV A ) is unique in comparison to the three previously characterized LpxJ homologs, further expanding the diversity of LpxJ family acyltransferase activity.
Our comparative genomics analysis indicates that LpxJ homologs (all members of DUF374) are widespread across Bacteria (Fig. 5A). These enzymes have functions analogous to the function of the canonical LpxM acyltransferase, and their discovery fills a long-standing hole in the Raetz pathway of lipid A biosynthesis in LpxM-deficient bacteria. Indeed, in silico structural prediction of LpxJ Rt indicates superficial similarities across LpxJ and PatA, a recently crystallized member of the LPLAT superfamily that contains an active site deviating from sites of canonical lipid acyltransferases (Fig. 5C). Further, we identified six highly conserved residues in all LpxJ homologs, which we confirmed were necessary for LpxJ function (Fig. 5A and B). These include a histidine at position 61 and an aspartate at position 132 that we believe represent an active-site catalytic dyad similar to PatA. Although these amino acids are distant in the primary sequence, structural modeling predicts a close spatial arrangement of these two residues (37, 41) (Fig. 5C). It is likely that these residues are positioned sufficiently close to allow charge interactions, abstracting a proton from the acceptor hydroxyl and facilitating nucleophilic attack of the incoming acyl chain thioester bond. This putative charge relay system is supported by the recently published crystal structures of PatA from Mycobacterium smegmatis (37,41), which does not contain the classical HX 4 D motif that is found in other lipid A acyltransferases (42)(43)(44) but is absent in LpxJ. Based on the conservation of the proposed catalytic residues within DUF374-containing acyltransferases, this active-site orientation might be common within the LpxJ family.
Lipid A architecture is fundamental to bacterial OM integrity. In many Gram-negative pathogens, changes in lipid A structure can have a profound impact on virulence (45,46). Therefore, we infer that Rickettsia LpxJ is also important in maintaining ideal membrane dynamics and facilitating molecular interactions at the host-pathogen interface that are required for adhesion and invasion of mammalian cells. Additionally, lipid A is the endotoxic component of LPS (47), and activation of TLR4 is critical for bacterial clearance in mouse models of Rickettsia infection (27,48), as well as other intracellular parasites (49)(50)(51). These data, as well as a recent report highlighting host-specific differences in lipid A gene expression (52), suggest a role for LpxJ in Rickettsia virulence, making this enzyme a tempting target for mutational studies in vivo. Further work promises to reveal novel insights into Rickettsia pathogenesis and contribute greatly to our understanding of rickettsial OM physiology.

MATERIALS AND METHODS
Recombinant DNA techniques, bacterial strains, and growth conditions. Primers used in this study were obtained from Integrated DNA Technologies (IDT) and are listed in Table 2. Genomic DNA was isolated from R. typhi grown in Vero76 tissue monolayer as previously described (53). lpxJ Rt (RT0047; NCBI accession number AAU03534) was amplified using Q5 polymerase 2ϫ master mix (M0492; NEB) and gel purified (740609.50; Macherey-Nagel) before infusion cloning (639649; Clontech) into a lac-inducible pFLAG-ctc vector (discontinued; Sigma) using HindIII and XhoI restriction sites. Expression of rickettsial genes after induction was verified by immunoblot detection of the C-terminal FLAG epitope tag. E. coli cultures in mid-log growth phase were induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) for 90 min at 42°C with shaking (225 rpm). Induced cultures and uninduced control cultures were harvested by centrifugation at 4,895 ϫ g for 15 min and suspended in 1ϫ NuPAGE Bolt sample buffer containing 1ϫ NuPAGE reducing agent at a concentration of 100 l of sample buffer for every 1 ml of culture. Samples were heated to 70°C for 10 min and sonicated briefly to fragment genomic DNA. Proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes, where recombinant LpxJ proteins were detected by indirect chemiluminescent immunoblot analysis using anti-FLAG (F1804; Sigma) primary antibody at a 1:5,000 dilution, followed by goat anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibody (405306; BioLegend) (see Fig. S3 in the supplemental material). The resulting constructs are designated pFLAG-lpxJ Rt and pFLAG-lpxJ Rr . Sitedirected mutagenesis reactions of selected residues within pFLAG-lpxJ Rt were carried out using a QuikChange Lightning kit (210519; Agilent). Primers for mutagenesis are listed in Table 2, and protein expression was verified as described above. Bacterial strains used in this study are listed in Table 3. From frozen stocks, E. coli was diluted 1:100 after overnight growth into LB broth supplemented with 1 mM magnesium and containing the appropriate antibiotics. Diluted cultures were then grown with shaking (225 rpm) at 37°C to mid-log phase. Protein expression was induced by addition of IPTG to a final concentration of 1 mM. Induction was carried out at 42°C with shaking (60 rpm) for 90 min. Whole cultures were harvested by centrifugation at 5,000 rpm for 10 min, the supernatants were discarded, and the pellets were either snap-frozen in liquid nitrogen before lipid A microextraction or lyophilized prior to fatty acid analysis.
Lipid A microextraction. Microextraction of lipid A from E. coli cultures was performed as previously described (54,55). Briefly, pellets from 5 ml of mid-log phase E. coli, grown and induced as described above, were extracted in 400 l of a solution containing five parts of isobutyric acid and three parts of 1 M ammonium hydroxide and heated at 100°C for 1 h, followed by a 15-min incubation on ice and centrifugation at 2,000 ϫ g for 15 min. Supernatant was collected and mixed in equal parts with water and then frozen and lyophilized. Contaminants were washed from the dried material by two rounds of methanol washes using 1 ml of methanol, followed by sonication and pelleting at 10,000 ϫ g for 5 min. The final product was reconstituted in 2:1:0.25 chloroform-methanol-water (50 l) along with 4 to 8 grains of Dowex ion exchange resin (Fisher Scientific, Pittsburgh, PA) and incubated with vortexing for at least 5 min. Solubilized lipid A molecules (1 to 2 l) were spotted onto a stainless steel target plate along with 1 l of Norharmane matrix (10 mg/ml in 2:1 chloroform-methanol) for MALDI analysis on a Bruker MicroFlex matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry instrument in negative-ion mode calibrated with Agilent tuning mix (G2421A; Santa Clara, CA), and data were processed using flexAnalysis software (Bruker Daltonics). All microextraction chemicals were obtained from Sigma-Aldrich unless otherwise noted.
GC fatty acid analysis. LPS fatty acids were converted to fatty acid methyl esters (FAMEs) and analyzed using gas chromatography flame ionization detection (GC-FID) as previously described (56). Briefly, lyophilized bacterial cell pellets from 50-ml cultures prepared as described above were incubated at 70°C for 1 h in 500 l of 90% phenol and 500 l of water. Samples were then cooled on ice for 5 min and centrifuged at 9,391 ϫ g for 10 min. The aqueous layer was collected, and 500 l of water was added to the lower (organic) layer and incubated again. This process was repeated two additional times, and all aqueous layers were pooled. Two milliliters of diethyl ether (E-138-1; Fisher) was added to the harvested aqueous layers. This mixture was then vortexed and centrifuged at 2,095 ϫ g for 5 min. The upper (organic) phase was then aspirated off, and 2 ml of diethyl ether was added back to the remaining aqueous phase. The mixture was vortexed and centrifuged at 2,095 ϫ g for 5 min, and the lower (aqueous) phase was collected and then frozen and lyophilized overnight. LPS fatty acids were converted to fatty methyl esters in the presence of 10 g of pentadecanoic acid (P-6125; Sigma) as an internal standard, using 200 l of 2 M methanolic HCl (Alltech, Lexington, KY) at 90°C for 18 h. Samples were cooled to room temperature, and 200 l of NaCl-saturated water was added. Converted fatty methyl esters were then extracted twice with hexane and run on an HP 5890 series 2 gas chromatograph. Retention times were correlated to fatty acids using GC-BAME standards (1114; Matreya, Pleasant Gap, PA).
In silico analysis. To evaluate Rickettsia typhi RT0047 (NCBI locus tag AAU03534) as an LpxJ homolog, we aligned it with previously characterized LpxJ proteins: Helicobacter pylori (AAD05834), Campylobacter jejuni (EAQ73466), and Wolinella succinogenes (CAE10797). To further assess conserved regions of LpxJ proteins, we retrieved 2,842 putative homologs from the NCBI nonredundant protein database in blastp searches using RT0047 as the query. Proteins were aligned, with four conserved regions further evaluated for conservation using WebLogo (57). Both above-mentioned multiple sequence alignments were