Lipoprotein N-Acylation in Staphylococcus aureus Is Catalyzed by a Two-Component Acyl Transferase System

Although it has long been known that S. aureus forms triacylated Lpps, a lack of homologs to known N-acylation genes found in Gram-negative bacteria has until now precluded identification of the genes responsible for this Lpp modification. Here, we demonstrate N-terminal Lpp acylation and chemotype conversion to the tri-acylated state is directed by a unique acyl transferase system encoded by two noncontiguous staphylococci genes (lnsAB). Since triacylated Lpps stimulate TLR2 more weakly than their diacylated counterparts, Lpp N-acylation is an important TLR2 immunoevasion factor for determining tolerance or nontolerance in niches such as in the skin microbiota. The discovery of the LnsAB system expands the known diversity of Lpp biosynthesis pathways and acyl transfer biochemistry in bacteria, advances our understanding of Lpp structural heterogeneity, and helps differentiate commensal and noncommensal microbiota.

inducible lit2 paralog that induces a weaker TLR2 response when grown in coppersupplemented media (21). Among staphylococci, Staphylococcus carnosus forms N-acetylated Lpps and induces 10-fold higher levels of the proinflammatory IL-6 cytokine than does S. aureus with TA-Lpp (22,23). Differences in immunostimulation among Lpp chemotypes thus helps to define the potential for virulence, as well as to facilitate the niche-specific adaptation of commensal bacteria from closely related noncommensal isolates (23).
Many of the enzymes directing Lpp N-terminal tailoring reactions, which in turn can modulate host TLR2 responses, remain to be discovered. Of particular note, S. aureus synthesizes TA-Lpp despite lacking an apparent ortholog to the N-acyl transferase (Lnt) used in Gram-negative bacteria (24)(25)(26). Schneewind and coworkers first reported on a base-stable Lpp acyl group, suggesting the presence of an amide-linked N-acyl chain (27). Analysis by mass spectrometry subsequently followed and provided direct structural evidence that unequivocally confirmed the TA-Lpp chemotype in S. aureus (28,29). Here, we report the identification of two previously uncharacterized genes (SAOUHSC_00822 and SAOUHSC_02761) in S. aureus required for this Lpp N-acylation using a TLR2/1-specific reporter assay to screen a random transposon library. We named this novel two-component Lpp tailoring machinery the Lipoprotein N-acyl transferase system (LnsA and LnsB). Neither LnsA or LnsB share any sequence similarity to Lnt or Lit, and both are only present in Staphylococcus species thus far known to make TA-Lpp. We show that loss of either gene in S. aureus converts TA-Lpp to DA-Lpp and that both genes are absolutely required and together sufficient to convert DA-Lpp to TA-Lpp in the L. monocytogenes. In either host, TLR2 challenge with DA-Lpp-producing strains induced a more potent response relative to isogenic TA-Lpp counterparts. Deletion of either lnsA or lnsB increased interleukin-8 (IL-8) secretion nearly 10-fold, indicating LnsAB are important determinants acquired by commensal and opportunistic staphylococci pathogens to evade TLR2 immune surveillance. The discovery of the presumptive twocomponent LnsAB complex expands the known diversity of Lpp tailoring reactions in bacteria.

RESULTS
Transposon library screen for diminished TLR2/1 signaling. We initially hypothesized that a single integral membrane protein in Staphylococcus aureus performs an analogous function to Lnt from Escherichia coli. Our selection strategy using growth rescue of an E. coli Lnt-depletion strain, however, that had successfully identified lit from Enterococcus faecalis and Bacillus cereus genomic DNA libraries (20) repeatedly failed when using S. aureus genomic DNA as the input library (data not shown). We thus turned to an indirect phenotypic screen to monitor loss of TLR2/1 signal, which has significantly higher affinity for TA-Lpp than DA-Lpp ligand, paired with a TLR2/6 specific activity counterscreen to eliminate candidates expressing either less total Lpp or that grew to a lower final biomass. Colonies from a high-coverage S. aureus mariner transposon (Tn) library built in the model lab strain NCTC8325 were used in the TLR2 reporter assays (30). We screened ϳ4,000 Tn mutants, and identified two unique Tn insertions in proximity to the unidentified open reading frame SAOUHSC_02761 (see Fig. S1 in the supplemental material). SAOUHSC_02761 is predicted to be a polytopic integral membrane protein by TMHMM2.0 (31), and has no sequence similarity with functionally annotated conserved domains. Remote protein homology with CAAX protein proteases (32) can be detected using HHPred, an algorithm that considers structure, as well as sequencing, to identify distant homology (33). The first Tn mutant (Tn 16C2) inserted 18 bp upstream from the SAOUHSC_02761 start codon, while the second (Tn 32F1) disrupted the coding region (amino acid 114 of a 249-amino-acid SAOUHSC_02761 ORF) (Fig. S1A). Northern blotting using an antisense SAOUHSC_02716 N-terminal probe confirmed a monocistronic transcript, that Tn 16C2 prevents readthrough expression, and that 32F1 expresses a longer transcript that is predicted to be frameshifted (Fig. S1B). Both Tn mutants had markedly diminished TLR2/1-specific stimulating activity comparable to disruption of the other known Lpp biosynthetic pathway enzymes, lipoprotein diacylglycerol transferase (Lgt) and lipoprotein signal peptidase II (Lsp) (Fig. S1C). Lgt links diacylglycerol through a thioether bond using a neighboring phospholipid to make preapolipoprotein on the extracellular membrane surface (34,35), which Lsp then cleaves to liberate the free ␣-amino cysteine terminus and complete DA-Lpp formation (36). Unlike Tn insertion in lgt and lsp, both Tn 16C2 and 32F1 mutants retained TLR2/6 activity at least as active as wild type (Fig. S1D). This is consistent with a common loss of SAOUHSC_02716 function genotype for both Tn mutants.
Although the immunoassay data implicated SAOUHSC_02761 in TA-Lpp formation, introducing SAOUHSC_02761 into an E. coli Lnt depletion strain failed to rescue growth (data not shown). We presumed an issue with heterologous expression or the available fatty acid donor pool in E. coli, so we repeated the TLR2 immunoassay using L. monocytogenes. Both S. aureus and L. monocytogenes have saturated branched-chain fatty acids (37,38), utilize acyl-phosphate donors in glycerophospholipid biosynthesis (39,40), and in general share much cell envelope physiology. However, introduction of SAOUHSC_02761 once again failed to induce phenotypic conversion (see below), indicating SAOUHSC_02761 may be required for TA-Lpp formation but not alone sufficient. We took advantage of the prearrayed Nebraska Transposon Mutant Library (NTML) in the S. aureus JE2 USA300 clinical isolate to repeat the TLR2/1 specific immunoactivity screen (Fig. 1A). We identified six Tn gene disruption mutants with decreased activity, including in SAUSA300_2405 [Tn NE407(Tn2405) or SAOUHSC_ 02761 in strain NCTC8325], lgt, and lsp. The only Tn library mutant besides SAUSA300_2405 that retained TLR2/6 activity while growing to a normal final biomass had an insertion in the uncharacterized open reading frame SAUSA300_0780 [Tn NE536(Tn0780) or SAOUHSC_00822 in strain NCTC8325]. The SAUSA300_0780/ SAOUHSC_00822 gene encodes a 189-amino-acid protein containing a domain with very weak similarity to the NlpC/P60 endopeptidase superfamily (41) and is predicted by SignalP v5.0 to contain a signal peptide for extracellular transport (42). As with SAUSA300_2405/SAOUHSC_02761, this gene is expressed under standard culture conditions but is part of a polycistronic operon as judged by transcript length (Fig. S1B). Expression levels of SAUSA300_0780/SAOUHSC_00822 were constant and not reliant on SAUSA300_2405/SAOUHSC_02761 function. Tn insertion in either SAUSA300_0780 or SAUSA300_2405 in S. aureus USA300 JE2 decreased TLR2/1 activation by Ͼ50-fold compared to the wild type (Fig. 1B).
To confirm we had identified all candidate genes required for synthesis of a TLR2/1-active Lpp chemotype, we heterologously expressed both SAOUHSC_00822 and SAOUHSC_0276 in L. monocytogenes. L. monocytogenes normally makes DA-Lpp, so introduction of both candidate genes from S. aureus should impart TLR2/1 activity. While either gene alone did not produce TLR2/1 signal, the expression of both genes with a constitutive promoter markedly enhanced the TLR2/1 response (Fig. 1C). Although other genes (such as those providing the acyl donor) may be required for Lpp chemotype conversion, these genes are evidently not specific to S. aureus. Deletion of either SAOUHSC_00822 or SAOUHSC_02761 decreases detection by TLR2/1 while enhancing TLR2/6 activity. To determine whether the decrease in TLR2/1 signal was solely due to Tn disruption and whether the two mutations are additive, we constructed targeted in-frame deletions of both genes in the S. aureus NCTC8325 background and measured TLR2/1 and TLR2/6 activity (Fig. 2). All deletion strains grew at rates statistically identical to the parent wild-type strain. Dilution series of heat-killed bacterial culture extracts were applied to TLR2/1-expressing reporter cells, and transcriptional activation of NF-B was measured. Deletion of either or both genes in tandem decreased the signal equivalently, which could be restored to near-wild-type levels by plasmid back-complementation ( Fig. 2A). With TLR2/6 assays, activity increased to the same extent in the single and double gene deletion constructs (Fig. 2B). The opposing TLR2 response indicates SAOUHSC_00822 and SAOUHSC_02761 are mutually required to swap TLR2 receptor ligand specificity. Since any combination of gene deletion alleles elicited equal changes in the TLR2 response, SAOUHSC_00822 and SAOUHSC_02761 are therefore codependent and not separate components of redundant Lpp N-acylation pathways.
SAOUHSC_00822 and SAOUHSC_02761 alter the Lpp profile in S. aureus. To correlate changes in TLR2 immunoassay data ( Fig. 2) with Lpp structure, we introduced a plasmid encoding a fragment of the S. aureus SitC Lpp with a C-terminal strep tag probe under the control of the constitutive promoter P tuf (Fig. 3A). The probe contained an N-terminal signal peptide for export, a lipobox for recognition by Lgt and Lsp for maturation, and the first 9 amino acids of SitC following the N-terminal cysteine linked to the strep tag epitope. The short probe length allowed separation of Lpp chemotypes with small mass differences by SDS-PAGE, including those due to total number of acyl chains. All S. aureus Lpp extracts produced a single homogenous band (Fig. 3B). In comparison to the wild type, the ΔSAOUHSC_00822 and ΔSAOUHSC_02761 constructs produced fastermigrating Lpp chemotypes that could be reverted back to wild type by plasmid The TLR2/1 specific activity was measured from heat-killed extracts from wild-type L. monocytogenes (expressing DA-Lpp) and isogenic strains expressing SAOUHSC_02761 (same as SAUSA300_2405, p2761), SAOUHSC_00822 (same as SAUSA300_0780, p822), or both genes together (p2761, p822). Statistical significances were calculated by using Student t tests (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001; ns, not significant).
LnsAB N-Acylates Lipoproteins in S. aureus ® back-complementation. Deletion of both genes (ΔSAOUHSC_00822 ΔSAOUHSC_ 02761) did not further change the Lpp profile, confirming a mutually nonredundant role for these genes in Lpp modification. To determine whether one candidate gene was needed for transcription of the other, we repeated the assay using a constitutive promoter (P pen ). The extent of complementation was complete and identical to vectors with native promoters. The functional codependence of SAOUHSC_00822 and SAOUHSC_02761 is thus not based on transcriptional regulation in line with initial Northern blotting results (Fig. S1B).
SAOUHSC_00822 and SAOUHSC_02761 constitute a novel lipoprotein N-acylation system (LnsAB) directing TA-Lpp synthesis in S. aureus. To determine whether the mass shift observed by SDS-PAGE was indeed due to acylation state, native Lpps were extracted and analyzed by matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (20,43). The N-terminal lipopeptide spectrum from wild-type SitC yielded a characteristic series of ions differing by 14 U (-CH 2 -), consistent with a highly heterogeneous population of lipopeptides varying in total acyl chain (Fig. 4). The majority of the total signal could be assigned to the TA-Lpp chemotype, with the C51 chemotype being the dominant ion (MϩH ϩ 1,353.06 U). Fragmentation . Heat-killed bacterial extracts were applied either as concentrated (black) or 5-fold diluted (gray) aliquots. (B) TLR2/6 receptor activity using the same extracts as in panel A, except that concentrated (black) or 10-fold-diluted (gray) aliquots were used. Error bars in both panels represent standard deviation results of at least three experimental replicates. Statistical significances (listed for black and gray bars) were calculated by using Student t tests (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001). of the C51 sodiated parent ion adduct (MϩNa ϩ 1,375.06 U) produced a series of N-acyl dehydroalanyl signals separated by two methylene units (28 U), with C 18:0 being the most abundant N-acyl fatty acid substitution (Fig. S2A). In contrast, deletion of SAOUHSC_00822 or SAOUHSC_02761 yielded MS spectra containing lower molecular mass lipopeptide signals between 1,057 and 1,137 U ( Fig. 4D and E). The most abundant signals were C33 and C35 Lpp chemotypes (MϩH ϩ 1086.7 U and 1114 u), nominally consistent with two acyl chains. Fragmentation of the C33 sodiated adduct (MϩH ϩ 1,108.7 U) confirmed the canonical DA-Lpp acyl chain distribution with a diacyl glycerol moiety and a free ␣-amino terminal cysteine (Fig. S2B). Deletion of both genes resulted in an identical MS profile with similar acyl chain composition to single gene deletion mutants (Fig. 4F). TA-Lpp synthesis could be restored in all constructs by plasmid back-complementation (Fig. S3). Analysis of a set of extracts from a different Lpp (SAOUHSC_02699) yielded identical results as with SitC (Fig. S4). The MALDI-TOF N-terminal Lpp modification attenuates detection by TLR2/1/6. With the discovery of LnsAB, we could now directly compare the TLR2-stimulating potential of DA-Lpp, TA-Lpp, and lyso-Lpp chemotypes in the same S. aureus genetic background. Reporter cells expressing TLR2/1/6 (capable of binding both DA-Lpp and TA-Lpp ligand) were challenged with heat-killed bacterial cells, and the NF-B transcriptional activation was measured (Fig. 5A). A clear hierarchy was observed in the bacterial cell count needed to reach half-maximal activation (EC 50 ). The EC 50 of S. aureus expressing DA-Lpp was Ͼ100-fold lower in comparison to an isogenic lyso-Lpp forming strain carrying lit from B. cereus. The EC 50 for wild-type S. aureus synthesizing TA-Lpp was intermediate, increasing the EC 50 from DA-Lpp by 10-fold. Lpp chemotype potency for TLR2 activation is hence ordered (DA-Lpp Ͼ TA-Lpp Ͼ lyso-Lpp), demonstrating how Firmicutes Lpp N-terminal modification systems can alter TLR2 detection over more than 2 orders of magnitude.
In order to translate changes in the NF-B transcriptional response into proinflammatory cytokine secretion, we repeated the assay using lnsAB Tn insertion mutants in the contemporary clinical isolate S. aureus USA300 [NE536(Tn0780)-LnsA and NE407(Tn2405)-LnsB initially identified in Fig. 2A]. If both S. aureus genes direct TA-Lpp formation, the immune-stimulating activity in lnsAB deletion mutants making only DA-Lpp should be significantly increased as well (Fig. 4). Immune stimulation was monitored by the production of IL-8. The wild-type JE2 parent produced only about 500 pg/ml, whereas in the two Tn mutants NE536(lnsA::Tn) and NE407(lnsB::Tn) the IL-8 production was ϳ10 times higher (Fig. 5B). When back-complemented in either NE536(pCX-LnsA) and NE407(pCX-LnsB), the IL-8 production was decreased. Conversely, when both genes were coexpressed in JE2 (pCX-LnsA/LnsB), IL-8 production was even further decreased, indicating high-level N-acyl transferase activity further shifted the Lpp chemotype population in favor of the weaker TA-Lpp TLR2 agonist. The cytokine response is in complete agreement with the NF-B transcriptional activation data. It has previously been shown that N-acetyl Lpp induces higher immune stimulation than TA-Lpp in S. carnosus (23), and direct comparison of S. carnosus to DA-Lppproducing S. aureus lnsA or lnsB mutants demonstrated nearly equivalent activity. Thus, staphylococci forming N-acetyl or DA-Lpp are higher TLR2 activating agonists than their TA-LPP counterparts.

DISCUSSION
There is much naturally occurring structural diversity among Lpp chemotypes in Firmicutes (19). Unlike in Gram-negative bacteria, deletion of the core Lpp biosynthetic genes (Lgt and Lsp) in monoderm Gram-positive bacteria such as S. aureus induces a subtle phenotype in rich media (44), and a robust phenotype specifically attributable to the Lpp N-terminal acylation state has yet to be reported. We therefore utilized loss of TLR2/1 activation to identify bacterial mutants with changes in the Lpp N-terminal acylation state. We screened two separate S. aureus Tn insertion libraries and, surprisingly, identified two previously unknown genes necessary for TA-Lpp production that we have now named LnsA (SAOUHSC_00822) and LnsB (SAOUHSC_02761). Both lnsAB genes are absolutely required for TA-Lpp formation in S. aureus as determined by the Lpp SDS-PAGE profiles (Fig. 3) and MALDI-TOF mass spectrometry ( Fig. 4; see also Fig. S4). LnsAB share no similarity with the two other known Lpp N-acylating enzymes in either primary amino acid sequence or mechanism. The apolipoprotein N-acyl transferase (Lnt) in Gram-negative bacteria utilizes the sn-1 acyl chain of phosphatidylethanolamine as an N-acyl chain donor (45), while Lit intramolecularly transfers the sn-2 acyl group of DA-Lpp to the ␣-amino terminus to form lyso-Lpp (46). The acyl chain source for LnsAB is currently unknown. While other genes may be needed to make the acyl donor, these genetic determinants are not unique to S. aureus since TLR2/1 agonist activity was conferred to L. monocytogenes by integrating just lnsAB into the genome (Fig. 1C).
Aside from being absolutely required for TA-Lpp formation ( Fig. 3 and 4) and TLR2/1 specific detection in S. aureus (Fig. 1B, 2, and 5) or L. monocytogenes when heterogeneously expressed (Fig. 1C), the function of LnsB in Lpp N-acylation is much more speculative than LnsA. LnsB does have very weak similarity to the CAAX prenyl protease from the archaeal methanogen Methanococcus maripaludis (Rce1 [4CAD_C] [60], 14% identity) and the APH-1A subunit of human ␥-secretase (APH-1A [5A63_C] [61], 7% identity). Both of these integral membrane proteins are part of the CAAX protease and bacteriocin-processing enzyme (CPBP) family (62), a large class of enzymes encompassing metalloproteases and other integral membrane proteins with poorly defined cellular function. Many bacteria encode multiple CPBPs, with S. aureus NCTC8325 containing at least six other CPBP enzymes in addition to LnsB (62). Of these, MroQ is a suspected protease that processes auto-inducing peptide (63,64). There is phenotypic evidence for roles of four other staphylococcal CPBPs in maintaining cell envelope integrity (65) and in the expression of cell wall-attached surface proteins with YSIRK peptide signals (66). It is apparent from these studies, however, that the cellular functions are not entirely overlapping, and at least one (SAOUHSC_02611/LyrA/SpdC) is almost certainly not a protease since key catalytic residues are absent (67). A similar analysis of catalytic motifs in LnsB shows considerable divergence from all CPBPsubfamily signature motifs as well, particularly in motif 4 (Fig. S5A). The similarity of LnsB to CAAX proteases, albeit without conservation of catalytic residues, suggests the CPBP fold could have been coopted for a noncatalytic chaperone role analogous to that suggested for the APH-1 subunit of ␥-secretase (68). APH1 has low sequence conservation of CPBP catalytic residues like LnsB (Fig. 6A) and no standalone proteolytic activity. Instead, APH1 is proposed to present protein substrate to the presenilin protease subunit core for hydrolysis within the ␥-secretase complex (69). Whether LnsB physically associates with LnsA in a complex, contributes any catalytic residues to the active site, or interacts with Lpp substrates remains to be determined. Alternative models where LnsB makes a novel acyl donor that is used by LnsA or where LnsB is indirectly required to process or stabilize LnsA cannot be ruled out.
There are also substantial differences in gene content and arrangement between the S. aureus LnsB genomic locus and the corresponding positions in both S. epidermidis and S. carnosus genomes. Genomic synteny in the LnsB loci between the staphylococci strains is highly mosaic, suggesting possible species-specific recombination or even horizontal acquisition events (Fig. S5B). While S. aureus (TA-Lpp) has common flanking genes with S. epidermidis (TA-Lpp) and S. carnosus (N-acetyl Lpp) on only one side, they are inverted with respect to LnsB (Fig. S5B). Gene architecture in S. epidermidis is intermediate and shares certain features with both genomes. Curiously, an LnsB-like CPBP open reading frame in the same position is present in all three genomes, including in the N-acetyl Lpp forming S. carnosus genome (SCA_1941). The overall sequence similarity between all three CPBP proteins is much lower (25 to 28% identical) though, in comparison to neighboring genome segments, which is inconsistent with expectations for simple species-driven genetic drift. Once more, no homology can be detected between the CPBP open reading frames at the DNA level, suggesting functional divergence as well as a possibly independent origin for all three genes. As with LnsB from S. aureus, the S. epidermidis ortholog (SE_2027) does not have many of the essential CPBP catalytic motifs and thus likely functions in Lpp N-acylation as well (Fig. S5A). The S. carnosus CPBP gene (SCA_1941) in comparison has all the CBPB canonical signature residues, including a completely intact motif 4. The sequence divergence may reflect the difference in catalytic activity (Lpp N-acetylation versus N-acylation) or, more likely, that SCA_1941 is not a functional LnsB ortholog and that the seemingly conserved synteny is due to genome rearrangement events.
The identification of LnsAB expands the catalog of known TLR2 recognition factors for staphylococci (Fig. 5). The TLR2-stimulating potential between different Firmicutes, and even within the same species, can vary significantly (11). In S. aureus, differences in total Lpp gene content, capsular polysaccharide thickness, and autolysis rates can all attenuate ligand release and accessibility (70). Phenol soluble modulins produced by S. aureus are surfactant-like small peptides that enhance release of Lpp-loaded extracellular vesicle release and in turn alter TLR2 responses (71). Disparity also stems from TLR2 antagonizing factors. The lipoylated E2 subunit of the pyruvate dehydrogenase complex suppresses TLR2/1 activity (72), while levels of secreted lipases that degrade shed Lpp is subject to lysogenic bacteriophages (73). An additional factor is the Lpp chemotype itself, which can indirectly attenuate the TLR2 response by acting through any of the above mechanisms, or more simply by altering ligand specificity and/or affinity at the respective TLR2 receptor complexes as demonstrated here. It should be noted that in staphylococci there are no other compounds described that activate TLR2 and Lpp is the dominant immunobiologically active ligand (23,74).
Although TLR2 attenuation is clearly an outcome of chemotype conversion from DA-Lpp to TA-Lpp, there is likely an overarching selective pressure for Lpp N-modification that supersedes immune evasion. Noncommensal strains such as S. carnosus and environmental strains such as B. subtilis have Lpps modified with N-acetyl groups. Analogous N-acetyl amino terminal tailoring has even been reported in archaea (75). Some archaea express Lpp-like, membrane-associated proteins with a characteristic lipobox preceding an invariant cysteine residue as in bacteria, except that they are thought to be modified with diphytanoyl glycerol diether lipid (76). TLR2 immunomodulation is an unlikely motivation for N-terminal tailoring in any of these cases. One possible clue regarding a broader, universal selective pressure operating outside the host TLR2-bacterial niche is offered by the recent discovery of an Lit2 paralog in L. monocytogenes (21). The lit2 gene is embedded within a copper resistance operon on either a transposon or transmissible plasmid in select environmental isolates. Like chromosomally encoded Lit, Lit2 converts DA-Lpp to lyso-Lpp but is specifically induced by copper ions. It was suggested that Lpp N-acylation may help prevent copper coordination at the membrane surface, limiting its uptake or oxidative damage from copper-mediated redox cycling (21). In E. coli, copper exposure induces DA-Lpp accumulation and Lpp outer membrane trafficking defects (77), while intracellular copper accumulates in Lnt depletion strains (78). Copper as a selective pressure would help explain the genesis of novel Lpp acylation systems such as LnsAB and more broadly the Lpp chemotype heterogeneity observed in Firmicutes. Copper selective pressure has grown in step with environmental oxygenation levels that increase copper bioavailability (79). Contemporary selection for copper resistance determinants has also arisen from copper's widespread current use as an antimicrobial agent (80). Assuming that the core Lgt-Lsp pathway was initially established in prokaryotes and that selective pressure for Lpp N-acylation arose after establishment of the various lineages, different bacterial species would have subsequently acquired N-acylation systems independently from each other. A particularly intriguing theory proposes modern Firmicute monoderm lineages independently arose at multiple times from a common diderm ancestor through loss of genes directing the biogenesis of the second outer membrane (81,82). If a common ancestral Lpp N-acylation system was lost in tandem, multiple independent Lpp N-acylating gene reacquisition events would have followed. In either case, the Lpp N-acylation system diversity exemplified by LnsAB has provided a ready-made genetic reservoir to modulate TLR2-mediated immunodetection among Firmicutes.

MATERIALS AND METHODS
Bacterial strains and growth conditions. All E. coli strains were grown in lysogeny broth-Miller medium (LB), while S. aureus and L. monocytogenes strains were grown in tryptic soy broth (TSB) at 37°C in baffled flasks (3-to-1 flask to culture ratio) with continuous aeration at 250 rpm unless indicated otherwise. For cytokine production assays, S. aureus strains were cultivated in basic medium (BM; 1% soy peptone, 0.5% yeast extract, 0.5% NaCl, 0.1% glucose, and 0.1% K 2 HPO 4 [pH 7.2]) at 37°C under continuous shaking at 150 rpm. Antibiotic resistance markers were selected with carbenicillin (100 g/ ml), kanamycin (30 g/ml), spectinomycin (50 g/ml), chloramphenicol (20 g/ml in E. coli, 10 g/ml for plasmid or 5 g/ml for integrated marker in S. aureus, and 3 g/ml in L. monocytogenes), and erythromycin (5 g/ml) where appropriate. The S. aureus NCTC8325 HG003 transposon library, built in the derivative strain TM226 (30), and the Nebraska Transposon Mutant Library (NTML), built in S. aureus strain USA300 (83), were cultured in 96-well microplates with buffered TSB (50 mM HEPES [pH 7.4]) without shaking in a humidified environment at 37°C. For pET22-based vectors, gene expression was induced using 1 mM IPTG (isopropyl-␤-D-thiogalactopyranoside). For expression of xylose-inducible genes encoded in plasmid pCX30, BM medium was supplemented with 10 g/ml chloramphenicol, and glucose was substituted by 0.5% (wt/vol) xylose. Generation times were measured in 96-well microplates in TSB media incubated at 37°C. All strains and plasmids are listed in Table 1.
Construction of bacterial deletion strains and plasmids. Gene deletions were constructed using the temperature-sensitive shuttle vector pKFC in S. aureus, as previously described (84,85). Plasmids were assembled from two separate 1-kb DNA fragments flanking the gene targeted for deletion and obtained by PCR. About 10 coding triplets from both ends of the targeted gene were retained to create nonpolar, in-frame gene deletions. Fragments were assembled using the In-Fusion HD cloning kit (TaKaRa Bio) and transformed into restriction negative S. aureus RN4220 by electroporation. Plasmids were then isolated and also transformed into strains TM226 and JE2 for integration and outcross in these genetic backgrounds. Deletion alleles were confirmed by PCR using primers annealing outside the targeted locus (Fig. S6). Complementation plasmids expressing SAOUHSC_00822, SAOUHSC_02761, or both genes in tandem were built from fragments amplified from S. aureus NCTC8325 genomic DNA and cloned into pLI50 (86), pCN59 (87), or pET22 (Novagen) using the same method and verified by sequencing. Expression of S. aureus genes in L. monocytogenes was achieved by integration into the chromosome using the pPL2 integration vector (88). The xylose-inducible pCX30 based complementation vectors (89) were constructed by PCR amplifying the two genes (SAUSA300_0780 and SAUSA300_2405) from S. aureus USA300 genomic DNA. The PCR inserts were cloned into pCX30 linearized with BamHI and SmaI using Hi-Fi DNA Assembly Master Mix (New England Biolabs). The resulting plasmid was transformed into S. carnosus TM300 by electroporation. Plasmid-harboring colonies were picked and verified by DNA  (30) and (ii) the NTML built in S. aureus strain USA300 JE2 (83). For the TM226 transposon library, the glycerol stock Tn pool was diluted and streaked to single colonies on tryptic soy agar (TSA; 5 g/ml erythromycin) and incubated overnight at 37°C. The next morning, individual colonies were inoculated into buffered TSB (200 l/well with 5 g/ml erythromycin) distributed in 96-well plates. Controls were included on each microplate (S. aureus RN4220 wild type and medium only). Microplates were incubated for 18 h at 37°C without agitation before bacterial cultures were resuspended by pipetting up and down three times. Aliquots (20 to 50 l) were transferred to a 96-well PCR plate, and bacteria were heat killed by incubation at 58°C for 1 h. The NTML screen was conducted in an identical manner, except that the growth microplates were inoculated with 5 l of thawed glycerol stocks from the prearrayed library stock plate. Heat-killed bacterial extracts were stored at 4°C until use.
Human embryonic kidney 293 cells (HEK-Blue hTLR2-TLR1; Invivogen) with endogenous TLR1 and TLR6 deleted and stably transfected with TLR2, TLR1, and an NF-B responsive secreted alkaline phosphatase reporter gene were cultured as recommended by the manufacturer and recently described (90). On the day of the assay, ϳ70% confluent HEK-Blue hTLR2-TLR1 cells were washed with 1ϫ phosphate-buffered saline (PBS), harvested by centrifugation, counted, and diluted to the recommended final concentration of ϳ280,000 cells/ml in Dulbecco modified Eagle medium (DMEM) without selective antibiotics. To each well containing 190 l of cell culture medium, 10 l of the heat-killed bacterial extract was added. Microplates were then incubated at 37°C in a 5% CO 2 atmosphere for 44 h (for TLR2/1 assays), which pilot studies determined to be optimal for the largest dynamic assay range. Secreted alkaline phosphate was assayed as described previously (90), with minor modifications. Aliquots (20 l) of supernatant were removed and added to 180 l of QuantiBlue detection reagent (Invivogen), followed by incubation for 4 h before the absorbance was measured at 620 nm. Defined TLR2 Lpp ligands prepared from E. coli cells expressing either TA-Lpp (KA548) or DA-Lpp (KA775) were used as stimulation controls.
Primary Tn mutant hits were struck to single colonies and the decrease in TLR2/1 activity assay results confirmed. These samples were then tested for retention of TLR2/6 specific activity using HEK-Blue hTLR2-TLR6 cells (Invivogen) as described above except cells were stimulated for 20 h. All genotypes were checked and confirmed by PCR using primers targeting the lgt, lsp, and candidate N-acylation genes in the prearrayed NTML (Fig. S7). For the TM226 library (Fig. S1E), Tn insertion sites were mapped by inverse PCR of circularized gDNA fragments, as described elsewhere, except that Taq␣I (New England Biolabs) was used for DNA restriction (91).
TLR2 dose-response HEK-Blue reporter assays. The TLR2 stimulating activity of Lpp N-acylation mutants was assayed using HEK-Blue hTLR2 (TA-Lpp and DA-Lpp responsive), HEK-Blue hTLR2-TLR1 (TA-Lpp responsive), and HEK-Blue hTLR2-TLR6 (DA-Lpp responsive) cells cultured and assayed as described above. Serial dilutions of heat-killed bacterial extracts were prepared as described above except bacterial cultures were grown to mid-log-growth phase (optical density at 600 nm [OD 600 ] of 1.0 to 1.5) with aeration in 14-ml culture tubes to limit accumulation of DA-Lpp during stationary phase (92). CFU/ml were obtained by plating three different dilutions of cultures on TSA and enumerating colonies after overnight incubation.
Total RNA Northern blotting. Northern blots were performed using a NorthernMax kit (Ambion) according to the manufacturer's instructions. Briefly, 1.5 g of total RNA for each strain of S. aureus were separated on a 1% MOPS (morpholinepropanesulfonic acid)-formaldehyde-agarose gel and transferred to a BrightStar-Plus positively charged nylon membrane (Invitrogen) using a Whatman Nytran SuPerCharge TurboBlotter kit (GE Healthcare Life Sciences) for 3.5 h. Samples were crosslinked to the membrane by baking at 80°C for 20 min. Biotin-labeled RNA probes were synthesized from DNA with gene-T7-specific primer sets (see Table S1 in the supplemental material) using a MaxiScript T7 transcription kit (Thermo Fisher), including the optional DNase digestion and cleanup with NucAway spin columns (Invitrogen). Probes were added to 10 ng/ml in Ultrahyb ultrasensitive hybridization buffer (Invitrogen), followed by incubation at 72°C for 16 h. The membranes were washed as directed using a NorthernMax kit, with the two high-stringency washes performed at 68°C. RNA was visualized with a chemiluminescent nucleic acid detection kit (Thermo Fisher) according to the manufacturer's instructions. Immunoblotting for strep-tagged Lpp probe. A plasmid expressing a 10-amino-acid fragment of the SitC Lpp with a C-terminal strep epitope under the control of the strong constitutive promoter P tuf was constructed in the shuttle vector pLI50. The plasmid pLI50-sitC10AA was transformed into various RN4220 strains, and cultures were grown to early log phase (OD 600 ϭ 0.5). Bacterial pellets were obtained by centrifugation, washed once with PBS, and resuspended in buffer (10 mM Tris-HCl [pH 8.0]) containing 50 g/ml of lysostaphin. Samples were incubated for 15 min at 37°C before being quenched with 4ϫ SDS-PAGE loading buffer. Samples were then heated at 70°C for 15 min before being clarified by centrifugation (18,000 ϫ g, 5 min). Aliquots of supernatant were loaded onto an 18% Tris-tricine minigel and separated by electrophoresis using the Tris-tricine running buffer system (93). Protein was transferred to a nitrocellulose membrane (0.2 M) and developed with an HRP-anti-strep tag conjugate as instructed by the manufacturer (StrepMAB-Classic HRP conjugate; IBA Life Sciences).
MALDI-TOF mass spectrometry. Lpps were prepared for mass spectrometry as previously described (20,43). Briefly, Lpps were extracted using the Triton X-114 phase partitioning method, separated with a 10% SDS-PAGE gel, and transferred to a nitrocellulose membrane. Bands corresponding to S. aureus SitC (SAOUHSC_00634) and to a periplasmic binding protein type 2 family (SAOUHSC_02699) lpp were trypsinized overnight. After elution from the membranes, samples were mixed with ␣-cyano-4-hydroxycinnamic acid (␣-CHCA) matrix and analyzed on an Ultraflextreme (Bruker Daltonics) MALDI-TOF mass spectrometer in positive reflector mode. MS-MS spectra were acquired in Lift mode.
Cytokine release assay. The cultivation of HEK-Blue hTLR2 cells and bacterial preparation for the stimulation assay were performed as described previously (23). Cells were cultured in DMEM (Thermo Fisher) supplemented with 10% fetal bovine serum, 50 mg/liter Normocin (InvivoGen), and 1ϫ HEK-Blue Selection (InvivoGen) at 37°C with 5% CO 2 supplementation. HEK-Blue hTLR2 cells were seeded with 5 ϫ 10 4 cells/200 l/well into 96-well cell culture plates, followed by incubation at 37°C with 5% CO 2 for 24 h. Bacterial cells from overnight culture with antibiotics added according to plasmids being carried (Fig. S8) were harvested and washed three times with Dulbecco PBS (DPBS) before measuring the OD 578 in DPBS. To calculate bacterial dosage (MOI [multiplicity of infection]), bacteria were set to OD 578 of 1.0, which equals 1 ϫ 10 8 CFU/ml. The final bacterial dosage (MOI of 2) was suspended in 50 l of the HEK-Blue hTLR2 medium and added to the cultured HEK-Blue hTLR2 cells (total volume of medium, 200 l). Stimulation by these bacteria was carried out for 18 h before cellular supernatants were collected for cytokine assays. IL-8 secreted was measured by using an IL-8 human ELISA kit (Thermo Fisher) according to the manufacturer´s instruction.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.