Remodeling of Lipid A in Pseudomonas syringae pv. phaseolicola In Vitro

Pseudomonas species infect a variety of organisms, including mammals and plants. Mammalian pathogens of the Pseudomonas family modify their lipid A during host entry to evade immune responses and to create an effective barrier against different environments, for example by removal of primary acyl chains, addition of phosphoethanolamine (P-EtN) to primary phosphates, and hydroxylation of secondary acyl chains. For Pseudomonas syringae pv. phaseolicola (Pph) 1448A, an economically important pathogen of beans, we observed similar lipid A modifications by mass spectrometric analysis. Therefore, we investigated predicted proteomes of various plant-associated Pseudomonas spp. for putative lipid A-modifying proteins using the well-studied mammalian pathogen Pseudomonas aeruginosa as a reference. We generated isogenic mutant strains of candidate genes and analyzed their lipid A. We show that the function of PagL, LpxO, and EptA is generally conserved in Pph 1448A. PagL-mediated de-acylation occurs at the distal glucosamine, whereas LpxO hydroxylates the secondary acyl chain on the distal glucosamine. The addition of P-EtN catalyzed by EptA occurs at both phosphates of lipid A. Our study characterizes lipid A modifications in vitro and provides a useful set of mutant strains relevant for further functional studies on lipid A modifications in Pph 1448A.


Introduction
The Gram-negative plant pathogen Pseudomonas syringae infects a wide range of economically important crop species. Severe, worldwide disease outbreaks caused by P. syringae have long prompted researchers to study this pathogen. As a result, P. syringae is one of the best-studied plant pathogens for investigating molecular mechanisms of pathogenicity and pathogen-host interactions [1]. P. syringae pv. phaseolicola (Pph) causes a halo blight of common beans, which leads to major yield losses in China, Australia and developing countries [2][3][4].
A major characteristic of Gram-negative bacteria is the presence of lipopolysaccharide (LPS) in the cell envelope. The LPS structure can be divided into three parts: the Opolysaccharide (OPS), a core oligosaccharide, and the lipid A, which anchors the molecule in the outer leaflet of the outer cell membrane. While the OPS and the core oligosaccharide PagP is an acyltransferase initially identified in Salmonella and later in other bacteria including P. aeruginosa and E. coli [19,34,35]. It is located in the outer membrane where it catalyzes the palmitoylation of lipid A. Palmitate is transferred from the sn-1 position of phospholipids to the N-linked (R)-3-hydroxymyristate chain on the proximal glucosamine unit of lipid A in Salmonella [35]. PagL is located in the outer membrane as well and hydrolyzes an ester-linked acyl chain at the O-3 position of the distal glucosamine of lipid A in P. aeruginosa PAO1 [18,36]. Removal of an acyl chain increases membrane fluidity and confers resistance to CAMPs [37]. In Salmonella enterica, constitutive PagL expression enhances outer membrane vesicle formation, resulting in overvesiculation, whereas a pagL deletion strain exhibits decreased OMV secretion [38]. The dioxygenase LpxO, located at the inner membrane, catalyzes the hydroxylation of secondary acyl chains in P. aeruginosa PAO1. This leads to increased resistance to CAMPs and is required for full virulence. Hydroxylation-proficient bacteria usually carry one copy of lpxO in their genome [11,39]; however, laboratory-adapted and clinical isolates may have two copies of lpxO. These two hydroxylation events can be attributed to two different lpxO orthologs (lpxO1 and lpxO2); however, their functionality has not been assigned to a specific position [14].
While lipid A biosynthesis and modifications have been studied in detail in many animal and human pathogens, these pathways are still largely unknown in plant-associated bacteria, including economically important plant pathogens of the P. syringae species complex. Here, we investigated lipid A structure modifications and the underlying lipid A-modifying enzymes as well as their regulators in the bean pathogen Pph 1448A. We identified the orthologous remodeling genes pagL, lpxO, and eptA of P. aeruginosa PAO1 in Pph 1448A. Furthermore, we identified putative PhoP and PhoQ orthologs, that might be involved in regulation of lipid A modifications. We generated knockouts of all these orthologous genes in Pph 1448A and analyzed the impact on the lipid A structure using mass spectrometry. These analyses show that the lipid A remodeling functions of PagL, LpxO, and EptA are conserved in Pph 1448A. We also demonstrate that deletion of lipid A remodeling genes does not impair LPS formation and Pph 1448A growth kinetics in vitro.

Orthologues of Lipid A-Modifying Genes Can Be Found in Pseudomonas spp.
Lipid A biosynthesis as well as lipid A-modifying enzymes are well described in P. aeruginosa [11]. Since structural features and dynamic adaption of the lipid A moiety play a crucial role in the host colonization process of P. aeruginosa, we were interested in whether orthologous genes are present in different plant-associated Pseudomonas spp., which fulfill a similar role. Therefore, predicted proteomes of different publicly available Pseudomonas genomes were screened for lipid A biosynthesis and lipid A modification-related protein sequences by BLASTP analysis using the respective protein sequence from P. aeruginosa PAO1 as a reference (Figure 1).
In all analyzed Pseudomonas spp., proteins involved in lipid A biosynthesis (LpxA, LpxC, LpxD, LpxH, LpxB, LpxK, LpxL, and LpxM) were identified with sequence identities of at least 72% compared to P. aeruginosa PAO1, suggesting that the basic lipid A synthesis is conserved in Pseudomonas. In contrast, the proteins involved in lipid A modification processes (PagL, LpxO1, LpxO2, EptA, ArnT, PagP) differ considerably in their sequence identity from their orthologs in P. aeruginosa in all Pseudomonas spp. examined. Whereas putative orthologs of PagL were identified with a protein sequence identity ranging from 59% in P. fluorescens FR1 and Pph 1448A up to 67% in P. syringae pv. maculicola ES4326, PagPmediated palmitoylation seems to be an uncommon feature in plant-associated Pseudomonas spp. Of all Pseudomonas strains investigated, only P. fluorescens A506 and P. fuscovaginae seem to have putative orthologs of PagP with sequence identities of 65% and 67%, respectively. Interestingly, all Pseudomonas strains analyzed share high sequence identity with LpxO2 from P. aeruginosa PAO1, whereas only P. fluorescens, P. fuscovaginae, and P. putida have a second homolog, closely related to LpxO1. A closer look at both identified LpxO proteins in these three strains reveals two different predicted proteins, suggesting that those strains express two proteins for lipid A hydroxylation. related protein sequences by BLASTP analysis using the respective protein sequence from P. aeruginosa PAO1 as a reference ( Figure 1).

Figure 1.
Comparison of sequence identities of lipid A biosynthesis and lipid A-modifying proteins as well as proteins involved in the regulation of lipid A modifications in different Pseudomonas species. P. aeruginosa PAO1 protein sequences were used as a reference, e-value cutoff = 1 × 10 −4 . Results of the BLASTP analysis are provided in Table S1.
In all analyzed Pseudomonas spp., proteins involved in lipid A biosynthesis (LpxA, LpxC, LpxD, LpxH, LpxB, LpxK, LpxL, and LpxM) were identified with sequence identities of at least 72% compared to P. aeruginosa PAO1, suggesting that the basic lipid A synthesis is conserved in Pseudomonas. In contrast, the proteins involved in lipid A modification processes (PagL, LpxO1, LpxO2, EptA, ArnT, PagP) differ considerably in their sequence identity from their orthologs in P. aeruginosa in all Pseudomonas spp. examined. Whereas putative orthologs of PagL were identified with a protein sequence identity ranging from 59% in P. fluorescens FR1 and Pph 1448A up to 67% in P. syringae pv. maculicola ES4326, PagP-mediated palmitoylation seems to be an uncommon feature in plant-associated Pseudomonas spp. Of all Pseudomonas strains investigated, only P. fluorescens A506 and P. fuscovaginae seem to have putative orthologs of PagP with sequence identities of 65% and 67%, respectively. Interestingly, all Pseudomonas strains analyzed share high sequence identity with LpxO2 from P. aeruginosa PAO1, whereas only P. fluorescens, P. fuscovaginae, and P. putida have a second homolog, closely related to LpxO1. A closer look at both identified LpxO proteins in these three strains reveals two different predicted proteins, suggesting that those strains express two proteins for lipid A hydroxylation.
Phosphoethanolamine transferase-like sequences were identified in all Pseudomonas spp. except P. fluorescens A506 and P. fuscovaginae, in which the analysis only resulted in hits of approximately 29% sequence identity. This may suggest that neither P. fluorescens A506 nor P. fuscovaginae modify their lipid A with P-EtN. Indeed, varying P-EtN content has been described in lipid A preparations of different P. fluorescens strains with some Comparison of sequence identities of lipid A biosynthesis and lipid A-modifying proteins as well as proteins involved in the regulation of lipid A modifications in different Pseudomonas species. P. aeruginosa PAO1 protein sequences were used as a reference, e-value cutoff = 1 × 10 −4 . Results of the BLASTP analysis are provided in Table S1.
Phosphoethanolamine transferase-like sequences were identified in all Pseudomonas spp. except P. fluorescens A506 and P. fuscovaginae, in which the analysis only resulted in hits of approximately 29% sequence identity. This may suggest that neither P. fluorescens A506 nor P. fuscovaginae modify their lipid A with P-EtN. Indeed, varying P-EtN content has been described in lipid A preparations of different P. fluorescens strains with some strains lacking almost any P-EtN in the lipid A portion [47]. For ArnT, analysis of predicted proteomes of P. syringae pv. tomato DC3000, P. fluorescens FR1, and P. putida K2440 yielded hits with approximately 26% sequence identity, whereas the analysis of other Pseudomonas proteomes resulted in hits with identities of 60% or higher. No hit was obtained for an ArnT homolog in P. syringae pv. tomato T1 and P. cichorii JBC1.
Furthermore, predicted proteomes of different Pseudomonas spp. were searched for the presence of the regulatory TCS PhoP/PhoQ and PmrA/PmrB. PhoP/PhoQ appeared to be conserved among Pseudomonas spp., since a sequence identity of at least 83% for PhoP and 64% for PhoQ was determined across all analyzed strains. Analysis of PmrA and PmrB protein sequences yielded often ambiguous and inconclusive hits with low sequence identities. Thus, we could not identify candidates for PmrA and PmrB in the strains studied.

Mass Spectrometric Analysis of Pph 1448A Lipid A Reveals the Activity of PagL, EptA, and LpxO
We chose Pph 1448A, a well characterized and economically relevant plant pathogen, as a model organism for our study of structural modifications of lipid A. To experimentally demonstrate the function of the identified gene loci, we generated isogenic knockout strains of pagL (PSPPH_1001), lpxO (PSPPH_1567), and eptA (PSPPH_1546) in Pph 1448A and analyzed their lipid A by MS 1 and MS 2 experiments. In general, the observed lipid A pattern showed a high similarity between the aqueous phase and phenolic supernatant LPS preparations of a respective strain. For Pph 1448A WT, the mass spectrum of lipid A released from LPS of the aqueous phase is depicted in Figure 2a and the spectrum for lipid A generated from LPS of the phenolic supernatant is shown in Figure 3a.
For the lipid A of isogenic mutants, the spectrum comprising the higher number of lipid A species was selected in each case and depicted in comparison to the lipid A of the Pph 1448A WT strain from the respective phase. Thus, MS spectra of the lipid A preparations from LPS of the aqueous phase are shown for the ∆lpxO (Figure 2b Tables 1 and 2. They include penta-and hexaacylated lipid A species present as mono-and di-phosphorylated variants, as well as lipid A species with one or two P-EtN modifications. The chemical structures of most of the observed lipid A species are summarized in Figure 4. Table 1. Summary of calculated monoisotopic neutral masses of lipid A species identified in Pph 1448A including the assignment of specific positions of primary and secondary fatty acids as well as phosphate and phosphoethanolamine modifications. Experimentally determined monoisotopic neutral masses are summarized in Table 2. P = monophosphate, PP-EtN = diphosphate ethanolamine.   Table 1; molecular species observed for each strain are listed in Table 2.   Table 1; molecular species observed for each strain are listed in Table 2.  Table 1; molecular species observed for each strain are listed in Table 2.   Lipid A from Pph 1448A WT comprises as main species penta-and hexa-acylated, di-phosphorylated lipid A species at 1430.873 Da and 1601.004 Da, respectively, in line with the typical fatty acid composition of Pseudomonas spp. as mentioned in the introduction. In addition, a prominent portion of the penta-acyl lipid A is only monophosphorylated (1350.904 Da). For both penta-acyl species, the mono-P-EtN substituted (1473.914 Da/1553.881 Da) versions are present; for the di-phosphorylated penta-species the di-P-EtN substituted lipid A can be observed as well (1676.891 Da). The same can be seen for the di-phosphorylated hexa-acyl lipid A that can have one (1724.013 Da) or two (1847.022 Da) P-EtN moieties in addition. All these eight lipid A species are accompanied by a second species with a mass difference of ∆m = 15.995 Da, pointing to a single addition of a hydroxyl group (peaks labelled in red in Figure 2). This hydroxylation is LpxO-dependent (Figure 2b), in line with the proposed function of LpxO as a dioxygenase that hydroxylates secondary fatty acids. These data were further corroborated by GC/MS analysis of the hydroxy fatty acids present in the respective lipid A preparations. Whereas in Pph 1448A WT lipid A 3-OH-C10:0, 3-OH-C12:0, and 2-OH-C12:0 fatty acids were detected, the latter was absent from lipid A preparation of the ∆lpxO strain (data not shown). Comparative MS 2 analyses of the di-phosphorylated hexa-acyl (1601.004 Da and 1616.999 Da; Figure 5) and penta-acyl (1430.873 Da and 1446.867 Da; Figure 6) lipid A species proved the sole presence of the secondary 2-OH-C12:0 at the distal glucosamine. Notably, the minor lipid A species with calculated monoisotopic masses of 1418.833 Da and 1541.842 Da (P-EtN-modified species of the aforementioned molecule), respectively, can potentially originate from two isomeric molecules that are the result of two different pathways (Table 1), e.g., penta-acyl lipid A with a mass of 1418.833 Da can be synthesized by PagL-mediated removal of a 3-OH-C10:0 fatty acid from the hexa-acylated lipid A with a mass of 1588.964 Da. Alternatively, this penta-acyl lipid A is a likely intermediate of the lipid A biosynthesis carrying only one secondary fatty acid [48]. Table 2. Calculated and experimentally determined monoisotopic masses of lipid A species observed in MS 1 spectra shown in Figures 2 and 3. Annotation accuracy of chemical structures to mass measurements are stated as ∆ppm. For more detailed structural information see Table 1 and      The presence of penta-acylated lipid A species can be attributed almost completely to the action of PagL. In the mass spectrum of the lipid A preparation from the ∆pagL strain ( Figure 2c) only a small basal level of such species is observable. The comparative MS 2 analyses mentioned above enabled us to assign the position of PagL-mediated removal of a 3-OH-C10:0 fatty acid to the O-3 position of the distal glucosamine.
While Figure 2 shows the effects of knocking out genes encoding fatty acid chainmodifying enzymes, Figure 3 focuses on the modification of phosphate residues with P-EtN. The P-EtN-modified lipid A species observed in Pph 1448A WT (Figure 3a) are absent in the ∆eptA strain (Figure 3b). Notably, the major mono-phosphorylated penta-acyl lipid A species (1350.904 Da, 1366.899 Da) found in the Pph 1448A WT and the other mutant strains are only present in low abundance in this preparation. Interestingly, the mass spectrum obtained for lipid A of the ∆phoPQ strain (Figure 3c) is very similar to the Pph 1448A WT spectrum (Figure 3a). The fragmentation of P-EtN-modified lipid A in MS 2 experiments is not as indicative as for the di-phosphorylated species. Nonetheless, the mono-P-EtN modification seems to be possible on both the 1-P or the 4 -P, but with a preference for 1-P as exemplarily shown for the fragmentation of mono-P-EtN substituted, di-phosphorylated penta-acyl lipid A (1553.881 Da; Figure S2). MS 2 experiments on mono-phosphorylated penta-acyl lipid A (1350.904 Da; Figure S3) revealed that this species lacks the 1-P.
For independent verification of the mass spectrometric data and structural alterations, crude lipid A samples generated by a small-scale preparation [49] were analysed by MALDI-TOF ( Figure S4).

Lack of Lipid A-Modifying Enzymes Does Not Affect Growth and LPS Formation in Pph 1448A
To test whether alterations of the lipid A structure affect growth behavior, we cultivated isogenic derivatives of Pph 1448A in liquid medium. All mutant strains showed WT-like growth kinetics (Figure 7). Alterations of the lipid A structure might influence the overall LPS structure. Analysis of Pph 1448A LPS via SDS-PAGE showed a similar ladder-like pattern in the mutant strains and the WT ( Figure S5). Thus, knockout of lipid A-modifying genes seems not to have major effects on OPS synthesis, LPS size distribution or LPS levels. While Figure 2 shows the effects of knocking out genes encoding fatty acid chain modifying enzymes, Figure 3 focuses on the modification of phosphate residues with P EtN. The P-EtN-modified lipid A species observed in Pph 1448A WT (Figure 3a) are absen in the ΔeptA strain (Figure 3b). Notably, the major mono-phosphorylated penta-acyl lipid A species (1350.904 Da, 1366.899 Da) found in the Pph 1448A WT and the other mutan strains are only present in low abundance in this preparation. Interestingly, the mass spec trum obtained for lipid A of the ΔphoPQ strain (Figure 3c) is very similar to the Pph 1448A WT spectrum (Figure 3a). The fragmentation of P-EtN-modified lipid A in MS 2 experi ments is not as indicative as for the di-phosphorylated species. Nonetheless, the mono-P EtN modification seems to be possible on both the 1-P or the 4′-P, but with a preference for 1-P as exemplarily shown for the fragmentation of mono-P-EtN substituted, di-phos phorylated penta-acyl lipid A (1553.881 Da; Figure S2). MS 2 experiments on mono-phos phorylated penta-acyl lipid A (1350.904 Da; Figure S3) revealed that this species lacks the 1-P.
For independent verification of the mass spectrometric data and structural altera tions, crude lipid A samples generated by a small-scale preparation [49] were analysed by MALDI-TOF ( Figure S4).

Lack of Lipid A-Modifying Enzymes Does Not Affect Growth and LPS Formation in Pph 1448A
To test whether alterations of the lipid A structure affect growth behavior, we culti vated isogenic derivatives of Pph 1448A in liquid medium. All mutant strains showed WT like growth kinetics (Figure 7). Alterations of the lipid A structure might influence the overall LPS structure. Analysis of Pph 1448A LPS via SDS-PAGE showed a similar ladder like pattern in the mutant strains and the WT ( Figure S5). Thus, knockout of lipid A-mod ifying genes seems not to have major effects on OPS synthesis, LPS size distribution or LPS levels.  Starting with an OD 600 of 0.1, bacterial growth was monitored hourly over 24 h using a photometer at 600 nm wavelength. No differences in growth kinetics of the mutant strains ∆pagL, ∆phoPQ, ∆eptA, and ∆lpxO compared to the WT strain were observed. Each growth curve depicts pooled data from three independent biological replicates. Statistical analysis using multiple t-tests comparing the growth of mutant strains hourly to the WT Pph 1448A did not show significant differences (p > 0.05).

Discussion
This study demonstrates the ability of P. syringae pv. phaseolicola 1448A to remodel its lipid A and identifies the corresponding gene loci of PagL, LpxO, and EptA. We performed mass spectrometric analyses to pinpoint the structural modification for each studied gene. In contrast to the predicted sequences of proteins involved in lipid A biosynthesis, sequences of lipid A-modifying proteins vary to a higher degree in all the strains examined. While LpxO2-and PagL-like protein sequences were found in all strains, the occurrence of LpxO1, EptA, ArnT, and PagP differs among Pseudomonas spp.
Palmitoylation of lipid A species alters host innate immune responses, increases resistance to some antimicrobial peptides and facilitates immune evasion of P. aeruginosa during colonization of its human host [19]. Interestingly, homologs of P. aeruginosa PagP have only been found in P. fluorescens A506 and P. fuscovaginae SE-1, yet the role of PagP in these plant-associated bacteria remains unknown. The absence of lipid A palmitoylation in most plant-associated Pseudomonas spp. may suggest that this trait is not beneficial for plant colonization or that the energetic costs outweigh the benefits.
We could not identify L-Ara4N additions to the lipid A of Pph 1448A by mass spectrometry, although Pph 1448A seems to have a homolog of ArnT with 63% sequence identity to P. aeruginosa ArnT. Possibly, it may not be expressed under the cultivation conditions used, since ArnT is regulated by PmrA/PmrB in P. aeruginosa and induced upon sensing of low pH, high Mg 2+ , Fe 3+ or Al 3+ [17,36,50]. Alternatively, the C-terminal truncation of 16 amino acids as compared to P. aeruginosa ArnT may render it inactive ( Figure S6).
Lipid A analysis of the ∆lpxO strain revealed the presence of a lipid A species that lack hydroxylation of secondary fatty acids (Figure 2). Our mass spectrometric data displayed a similar structural phenotype as described for P. aeruginosa PAO1 lpxO-mutants by Lo Sciuto et al. [14]. In contrast to P. aeruginosa PAO1 with two LpxO enzymes, we identified an LpxO2 ortholog in Pph  Figures 5 and 6). Thus, PSPPH_1567 encodes the lipid A hydroxylase LpxO in Pph 1448A. Like PagL activity, LpxO activity is important for pathogenicity and confers resistance to CAMPs in P. aeruginosa [14]. Additionally, lpxO mutants in Acinetobacter baumanii are less virulent compared to strains expressing a fully hydroxylated lipid A moiety [39]. In Pph 1448A, the hydroxylation of lipid A could possibly also support the colonization of its plant host Phaseolus vulgaris.
In P. aeruginosa, the outer membrane lipase PagL hydrolyses the ester-linked acyl chain at position O-3 of lipid A, which results in the release of 3-OH-C10:0 [18]. While Pph 1448A synthesizes hexa-and penta-acylated lipid A (Figures 2a and 3a), Pph 1448A ∆pagL predominantly expresses hexa-acylated lipid A species (Figure 2c). Our results show that the function of PagL (PSPPH_1001) is conserved in Pph 1448A. Notably, in order to generate a full length deletion of pagL, the overlapping neighboring hypothetical open reading frame of unknown function (PSPPH_1002) was C-terminally truncated ( Figure S7). In Salmonella, removal of the position O-3 acyl chain confers resistance to CAMPs and lowers the affinity for the TLR4/MD2 receptor and enables Salmonella to evade immune responses in mammalian hosts [37,51,52]. It is unknown whether P. vulgaris recognizes the lipid A moiety of Pph 1448A as a microbe-associated molecular pattern (MAMP), which triggers immune responses, and if removal of an acyl chain leads to evasion of plant immunity. Medium-chain 3-hydroxy fatty acids, ranging from C8:0 to C12:0, from Pseudomonas trigger immune responses through the cell surface-localized immune receptor LORE in Arabidopsis thaliana and other crucifers [15,53]. Interestingly, many plant-associated Pseudomonas spp. have orthologs of PagL and presumably release 3-OH-C10:0 fatty acids.
However, apparently only Brassicaceae evolved a defense mechanism to recognize free medium-chain 3-hydroxy fatty acids [53].
Mass spectra from lipid A of Pph 1448A revealed single and double modified phosphate groups carrying P-EtN modifications. P-EtN modifications were not detected in the respective ∆eptA strain (Figure 3b), while all other expected non-hydroxylated and hydroxylated lipid A species were present. Hence, our results show that PSPPH_1546 encodes the phosphoethanolamine transferase EptA in Pph 1448A. Orthologous genes of eptA can be found in a wide variety of plant-associated Gram-negative bacteria ( Figure 1). P-EtN additions to the lipid A are mainly used to decrease the overall net negative charge of the molecule [54]. Subsequently, CAMPs are less attracted to the lipid A moiety with a less negative net charge. The overall critical threshold of CAMPs on the bacterial membrane surface is lower and the membrane does not collapse, resulting in the survival of bacteria [55]. EptA-mediated P-EtN additions in Pph 1448A could help the pathogen to colonize bean plants. Expression of defensin-like antimicrobial peptides against different bacteria has been shown for Vigna sesquipedalis [56], for example, and could explain the general importance for P-EtN additions to the lipid A in plant-associated Pseudomonas spp.
Release of OMVs can be linked to P-EtN additions in C. rodentium and PagL activity in Salmonella. In C. rodentium, the release of OMVs is negatively affected by the activity of the two P-EtN transferases EptA and CptA, whereas deacylation of lipid A leads to a decrease in the hydrophobic cross-section area of lipid A and promotes OMV release in Salmonella [33,38]. Pathogenic Gram-negative bacteria export virulence factors in OMVs to aid the colonization process of the host. However, OMVs also contain MAMPs, which trigger plant immunity and prepare the plant to an upcoming pathogen attack [57,58]. Taken together, Pph 1448A may regulate the release of OMV during host colonisation through the addition of P-EtN, and PagL might play a role in evading the immune system in plants but may also allow recognition of the pathogen by cruciferous hosts.
Mass spectrometric analysis of lipid A obtained from the WT and from its isogenic mutant strains ∆lpxO and ∆phoPQ showed significant amounts of dephosphorylated lipid A species (Figures 2 and 3). Gram-negative bacteria are known to dephosphorylate their lipid A via phosphatases. For example, Helicobacter pylori utilizes the lipid A 1-phosphatase LpxE and lipid A 4 -phosphatase LpxF to promote its resistance to antimicrobial peptides and secure its survival during host colonization [59,60]. A similar lipid A-modifying process might be utilized by Pph 1448A. However, dephosphorylated lipid A species were found when Pph 1448A was cultivated in a full-strength medium. Identification of dephosphorylation events during bean plant colonisation could shed light on the relevance of lipid A phosphatases for plant-associated bacteria. Unlike the lipid A of the Pph 1448A WT, we identified little to no dephosphorylated lipid A species in the ∆eptA mutant. In the absence of P-EtN in the outer membrane, the overall net charge was presumably balanced in this way and phosphatase activity was not necessary to rebalance P-EtN additions under the growth conditions used. Interplay between lipid A phosphatase activity and P-EtN additions may provide pathogens with a strategy to establish itself in an unfavourable environment.
In this study, the detailed characterization of different lipid A modifications ( Figure 8) using mass spectrometric analysis was realized, which aids in understanding lipid A remodeling processes in plant-associated Pseudomonas spp. Precise annotation of functional group positions on the lipid A moiety enables follow-up studies to examine regulatory processes and modification during host infection. Taken together, our detailed genetic and structural analyses demonstrate the potential of Pph 1448A as a model organism to gain insights into the role of lipid A modifications during pathogen-plant interactions.

Strains and Growth Conditions
Strains were grown under shaking at 28 °C for P. syringae pv. phaseolicola [61] (gift by John Mansfield, Imperial College London, UK) or 37 °C for E. coli strains. Strains us in this study are listed in Table S2. Bacteria were grown in King's B medium (KB) [62], Lysogeny Broth [63]. Antibiotics were used at a final concentration of gentamicin at 5 mL −1 , rifampicin at 50 µg mL −1 , kanamycin at 25 µg mL −1 . To identify positive E. coli clon X-gal and IPTG were added to growth media at 30 µg mL −1 and 0.05 mM, respectively.

Analysis of Predicted Proteomes of Pseudomonas spp.
Predicted proteomes of different Pseudomonas spp. were analyzed to determi amino acid sequence identity using the BLASTP function on pseudomonas.com (acce date 7 December 2021). Cut-off was set to 1 × 10 -4 , word size 3, filtered and no pairw output. If multiple hits were obtained in a single predicted proteome, the sequence ide tity of the best hit according to the expected cut-off value and the bit score were used a depicted in the Table S1 and Figure 1

Gene Knockout in Pph 1448A
Pph 1448A knockout mutants were generated using the pGGKO-blue plasmid as d scribed previously with minor changes [64]. In brief, flanking sequences (450 to 650 bp) target genes were amplified via PCR from genomic DNA and inserted into pGGKO-bl backbone using Golden-gate cloning. Competent E. coli DH5α [65] cells were transform with this precursor plasmid. Precursor plasmid was isolated and a Gm R resistance casse was inserted between the flanking sequences. Final plasmids were verified by Sanger quencing. Mutants of Pph 1448A were generated as described with minor changes [6

Strains and Growth Conditions
Strains were grown under shaking at 28 • C for P. syringae pv. phaseolicola [61] (gifted by John Mansfield, Imperial College London, UK) or 37 • C for E. coli strains. Strains used in this study are listed in Table S2. Bacteria were grown in King's B medium (KB) [62], or Lysogeny Broth [63]. Antibiotics were used at a final concentration of gentamicin at 5 µg mL −1 , rifampicin at 50 µg mL −1 , kanamycin at 25 µg mL −1 . To identify positive E. coli clones X-gal and IPTG were added to growth media at 30 µg mL −1 and 0.05 mM, respectively.

Analysis of Predicted Proteomes of Pseudomonas spp.
Predicted proteomes of different Pseudomonas spp. were analyzed to determine amino acid sequence identity using the BLASTP function on pseudomonas.com (access date 7 December 2021). Cut-off was set to 1 × 10 -4 , word size 3, filtered and no pairwise output. If multiple hits were obtained in a single predicted proteome, the sequence identity of the best hit according to the expected cut-off value and the bit score were used and depicted in the Table S1 and Figure 1

Gene Knockout in Pph 1448A
Pph 1448A knockout mutants were generated using the pGGKO-blue plasmid as described previously with minor changes [64]. In brief, flanking sequences (450 to 650 bp) of target genes were amplified via PCR from genomic DNA and inserted into pGGKO-blue backbone using Golden-gate cloning. Competent E. coli DH5α [65] cells were transformed with this precursor plasmid. Precursor plasmid was isolated and a Gm R resistance cassette was inserted between the flanking sequences. Final plasmids were verified by Sanger sequencing. Mutants of Pph 1448A were generated as described with minor changes [66]. Triparental mating conjugation with E. coli HB101 [67] as a helper strain was used to transfer the plasmids into Pph 1448A. Positive recombinants were counter selected on KB Rif Gent Sucrose and insertion of the Gm R resistance cassette verified by Sanger sequencing of the respective gene loci (Figures S7-S10). Oligonucleotides used in this study are listed in Table S3.

Preparation of Lipopolysaccharide from Pph 1448A Strains
Pph 1448A strains were cultivated overnight under shaking and harvested via centrifugation (4 • C, 8000× g, 20 min) at an OD 600 = 1.2 to 1.5. Bacterial pellets were washed two times with pre-cooled ddH 2 O prior to freeze drying. Freeze-dried bacterial pellets were dissolved in 100% EtOH and the suspension was stirred for 2 h at room temperature. The suspension was filtered (Whatman 595 1 2 ) and sequentially washed with acetone (twice) and diethyl ether. Washed pellets were dried and resuspended in water (15 mg mL −1 ). 10% NaN 3 was added to a final concentration of 0.02% to the dissolved pellets. Dissolved pellets were sequentially treated with DNase/RNase (10 mg mL −1 w/v) and proteinase K (10 mg mL −1 w/v) at room temperature (100 µL enzyme per g dry weight). Digested pellets were dialyzed (14 kDa cut-off) for 2 days in ddH 2 O and freeze dried.
For LPS isolation, the hot phenol-water method was used [68]. In brief, dried dialysates were dissolved in pre-warmed ddH 2 O (68 • C, 100 mL per 10 g dry weight) until dialysates were completely dissolved. 90% aqueous phenol (equal volume as ddH 2 O) was added and stirred for 30 min at 68 • C. Aqueous-phenol suspensions were centrifuged (5600× g, 4 • C, 20 min) and the upper aqueous phase was collected. The extraction was repeated with the same amount of ddH 2 O that had been collected. Combined aqueous and phenolic phases were dialyzed against deionized water separately. Phenolic phases were separated via centrifugation (600× g, 20 • C, 5 min) into supernatant and phenolic pellet before freeze-drying.

Lipid A Preparation
Lipid A was prepared from LPS of the aqueous phase and the phenolic supernatant as described [53]. Briefly, LPS was dissolved in water (7.5 mg mL −1 ) and a solution of 10% SDS (volume equivalent to 12.5% of the water volume) as well as the same volume of acetate buffer (1 M NaOAc, pH 4.4) were added. The mixture was heated for 3 h at 100 • C under slight stirring and freeze-dried. SDS was removed by four washes with 30 mL 2 M HCl/EtOH (1:99 v/v (6000× g for 20 min at 20 • C)). The dried pellet was resuspended in 4 mL water. Afterwards, 4 mL CHCl 3 and CH 3 OH (4:1 v/v) were added and the suspension was mixed vigorously and centrifuged (6000× g) for 10 min at 4 • C. The organic phase was collected and the water phase (including the interphase) was extracted again three times with 3 mL CHCl 3 . All organic phases were combined and dried under a stream of nitrogen.

Mass Spectrometric Analysis of Lipid A via ESI-MS
All mass spectrometric analyses of lipid A preparations were performed on a Q Exactive Plus (ThermoFisher Scientific, Bremen, Germany) using a Triversa Nanomate (Advion, Ithaca, NY, USA) as nano-ESI source. Lipid A extracts were initially dissolved in a concentration of 5 µg µL −1 in chloroform:methanol:water (60:30:4.5 v/v/v). 5 µL of this solution were mixed with 95 µL of water/propan-2-ol/7 M triethylamine/acetic acid (50:50:0.06:0.02 v/v/v/v). Mass spectra were recorded for 0.50 min in the negative mode in an m/z-range of 400-2500 applying a spray voltage of −1.1 kV. All depicted MS 1 spectra were charge deconvoluted (Xtract module of Xcalibur 3.1 software (ThermoFisher Scientific, Bremen, Germany)) and all provided values refer to the monoisotopic mass of neutral molecules.
To further investigate the lipid A structure, MS 2 experiments were performed using 5 µL of the above mentioned lipid A solutions mixed with 10 µL chloroform:methanol:water (60:30:4.5 v/v/v) and 285 µL water/propan-2-ol/30 mM ammonium acetate/acetic acid mixture (15:15:1:0.04 v/v/v/v). Shortly before analysis, 0.5 µL triethylamine were added and the mixture was thoroughly mixed. Single charged ions of interest were selected and spectra were recorded in a positive ion mode at normalized collision energies (NCE) of 1, 10, 12, 14, 16, 20, and 30. Under these ionization condition, intensive Et 3 N adduct ions are formed. Usually, the di-Et 3 N adducts were selected for MS 2 experiments, only for monophosphorylated lipid A, the mono-Et 3 N adduct was selected. Applying increased NCE leads to the formation of the [M+H] + ion as well as the abundant B-fragment (according to the nomenclature of [69]). At an NCE of 12 or 14, the corresponding Y-fragment ion can usually be well detected, especially for di-phosphorylated lipid A species. Depending on the individual fragmentation of the investigated lipid A species, MS 2 spectra of NCE values with highest structural information content were selected for representation. The general distribution of 3-OH-acyl chains in Pseudomonas lipid A molecules is assigned according to and in line with earlier published structural analysis [70].

Analysis of Fatty Acids via GC-MS
The nature of hydroxy fatty acids in lipid A preparations of Pph 1448A WT and ∆lpxO was determined by generating the trimethylsilyl (TMS) derivatives of the respective FAMEs in comparison to authentic standards. The gas-liquid chromatography-mass spectrometry (GLC-MS) analyses were performed on an Agilent Technologies 6890N gas chromatograph coupled to a 5975 inert XL Mass Selective Detector (Agilent Technologies Santa Clara, CA, USA). A 30-m Agilent J&W DB-WAX Ultra Inert column (0.25 mm inner diameter, 0.25 µm film thickness) was used and a temperature gradient starting at 70 • C (kept for 1.5 min), then raised at 60 • C/min to 150 • C, kept there for 5 min, and raised to 220 • C at 1.5 • C/min was applied.

Mass Spectrometric Analysis of Crude Lipid A Extracts via MALDI-TOF
Freeze-dried bacterial pellets were rehydrated with endotoxin-free water, vortexed, then pelleted. Supernatant was discarded. Lipid A was extracted from cell pellets using an ammonium hydroxide-isobutyric acid-based procedure [49,72]. Briefly, a bacterial cell pellet was resuspended in 400 µL of 70% isobutyric acid (Sigma-Aldrich, I1754-1L) and 1 M ammonium hydroxide (Sigma-Aldrich, St. Louis, MO, USA, 221228-500ML-A) (5:3 v/v). Samples were incubated for 1 h at 100 • C and centrifuged at 8000× g for 5 min. Supernatants were collected, added to endotoxin-free water (1:1 v/v), snap-frozen on dry ice, and lyophilized overnight. The resultant material was washed twice with 1 mL methanol (Fisher Scientific, Waltham, MA, USA, A456-1), and lipid A was extracted using 80 µL of a mixture of chloroform (Fisher Scientific, C606SK-4), methanol, and water (3:1:0.25 v/v/v). Once extracted, 1 µL of the concentrate was spotted on a steel re-usable MALDI plate (Hudson Surface Technology, Closter, NJ, USA, PL-PD-000040-P) followed by 1 µL of 10 mg mL −1 norharmane matrix (Sigma-Aldrich, NG252-1G) in chloroform-methanol (2:1 v/v) (Sigma-Aldrich, St. Louis, MO, USA) and was then air dried. All samples were analyzed on a Bruker Microflex mass spectrometer (Bruker Daltonics, Billerica, MA, USA) in the negative-ion mode with reflection mode. An electrospray tuning mix (Agilent Technologies, Foster City, CA, USA, G2421A) was used for mass calibration. Spectral data were analyzed with Bruker Daltonics FlexAnalysis software (v4.30). The resulting spectra were used to estimate the lipid A structures present in each strain based on their predicted structures and molecular weights.

SDS-PAGE and Silver Staining of Bacterial LPS
2.5 µL of crude LPS sample was mixed with 5 µL of NuPAGE™ LDS Sample Buffer (Invitrogen, Carlsbad, CA, USA, NP0007), NuPAGE™ Sample Reducing Agent (Invitrogen, NP0004) and 15.5 µL ddH 2 O. Samples were incubated at 70 • C for 10 min prior to loading on the gel. LPS samples were separated using NuPAGE™ 4 to 12%, Bis-Tris, 1.0 mm, Mini Protein Gel (Invitrogen, NP0321) with MES as running buffer at 200 V constant settings for 40 min. After separation, the gel was kept for 2 h in fixing solution (30% v/v, EtOH, 10% v/v acetic acid), transferred into oxidizing solution (7% v/v periodic acid, 30% v/v EtOH, 10% v/v acetic acid) and washed three times with ddH 2 O. The gel was incubated in dye solution (0.1% w/v AgNO 3 in ddH 2 O) in the dark for 30 min before revealing signals on the gel with the revealer solution (3% w/v Na 2 CO 3 , 0.02% v/v formaldehyde) for 10 min in the dark. Coloration was stopped with 1% v/v acetic acid aqueous solution; the gel was washed three times with ddH 2 O for 10 min and photographed. Three independent biological replicates were carried out, showing the same pattern on the gel.

Bacterial Growth Curve
Bacteria were grown overnight in KB supplemented with respective antibiotics. Starting OD 600 of main cultures was adjusted to 0.1 in KB and the samples incubated at 28 • C with shaking. OD 600 was monitored every hour over 24 h. Three independent biological replicates were carried out and statistical analysis was assessed using multiple t-tests (alpha = 0.05, Holm-Sidak method) comparing the bacterial growth of the mutant strains to the WT every hour. Data Availability Statement: All data supporting the findings of this study are provided in the manuscript and its Supplementary Files. Additional data supporting the findings of this study are available from the corresponding authors upon request.