Extracellular zinc induces phosphoethanolamine addition to Pseudomonas aeruginosa lipid A via the ColRS two-component system

Summary

Gram-negative bacteria survive harmful environmental stressors by modifying their outer membrane. Much of this protection is afforded upon remodeling of the lipid A region of the major surface molecule lipopolysaccharide (LPS). For example, the addition of cationic substituents, such as 4-amino-4-deoxy-L-arabinose (L-Ara4N) and phosphoehthanolamine (pEtN) at the lipid A phosphate groups is often induced in response to specific environmental flux stabilizing the outer membrane. The work herein represents the first report of pEtN addition to P. aeruginosa lipid A. We have identified the key pEtN transferase which we named EptA_Pa and characterized its strict activity on only one position of lipid A, contrasting from previously studied EptA enzymes. We further show that transcription of eptA_Pa is regulated by zinc via the ColRS two-component system instead of the PmrAB system responsible for eptA regulation in E. coli and S. enterica. Further, although L-Ara4N is readily added to the same position of lipid A as pEtN under certain environmental conditions, ColR specifically induces pEtN addition to lipid A in lieu of L-Ara4N when Zn²⁺ is present. The unique, specific regulation of eptA_Pa transcription and enzymatic activity described in this work demonstrates the tight yet inducible control over LPS modification in P. aeruginosa.

Introduction

Pseudomonas aeruginosa ubiquitously inhabits soil and water sources and is known for its intrinsic tolerance to potentially toxic contaminants such as heavy metals (Caille et al., 2007). It is also a formidable opportunistic pathogen frequently acquired in healthcare facilities due to its persistence in sinks, showers and many non-aquatic abiotic surfaces (Lyczak et al., 2000; Kerr and Snelling, 2009; Chai et al., 2014). Once inside a human host P. aeruginosa thrives in a variety of tissue types resulting in acute skin, eye, and burn wound infections (Lyczak et al., 2000). Perhaps most notably, however, are chronic P. aeruginosa infections that persist within the lungs of cystic fibrosis patients for years and are recalcitrant to most antimicrobial treatment (Kerr and Snelling, 2009; Moskowitz and Ernst, 2010).

The outer membrane of Gram-negative bacteria like P. aeruginosa acts as a protective barrier to prevent binding and uptake of toxic molecules. The major component of the outer leaflet, lipopolysaccharide, interfaces with the environment and is often remodeled to protect the cell from environmental stressors. Lipopolysaccharide is composed of three distinct domains: a lipid A anchor, a core sugar region, and an outer polysaccharide known as O-antigen (Whitfield and Trent, 2014). The lipid A domain is a potent immunostimulant, hence its other name, “endotoxin” (Needham et al., 2013). Furthermore, the negatively charged lipid A molecule is a prime target of cationic antimicrobial peptides that destabilize the outer membrane resulting in cell lysis (Vaara and Vaara, 1981; Needham and Trent, 2013).

Modification of the canonical, hexa-acylated, bis-phosphorylated lipid A molecule produced in Gram-negatives (Fig 1A; black) alter its chemical properties to bolster membrane integrity. A repertoire of modification enzymes is responsible for the dynamic structure of P. aeruginosa lipid A. Previous work from our laboratory recently revealed that P. aeruginosa has an LpxT lipid A kinase that adds an additional phosphate group to the 1- or the 4' position under standard laboratory growth conditions (Fig. 1A; brown)(Nowicki et al., 2014). Hydroxylation of the secondary acyl chains can also occur by one of two LpxO enzymes, although the purpose for this modification remains to be elucidated (Fig. 1B; orange) (King et al., 2009). In addition to these modifications, the toxicity of P. aeruginosa lipid A can be affected by altering the acylation pattern due to activity of the PagL deacylase or the PagP palmitoyltransferase (Fig. 1B; pink and green) (Ernst, 1999; Ernst et al., 2003; Ernst et al., 2006; Thaipisuttikul et al., 2014). Aside from influencing endotoxicity, lipid A modifications can contribute to antimicrobial peptide resistance (Needham and Trent, 2013). Addition of 4-amino-4-deoxy-L-arabinose (L-Ara4N) to either phosphate group of P. aeruginosa lipid A by the enzyme ArnT is one such strategy (Fig. 1B; blue)(Bhat et al., 1990; Fernández et al., 2013). Palmitoylation has also been shown to increase antimicrobial resistance (Thaipisuttikul et al., 2014).

Fig 1.

Lipid A structure of P. aeruginosa with and without inducible modifications. A) Canonical, hexa-acylated, bis-phosphorylated lipid A structure is shown in black. Phosphorylation of the lipid A phosphate groups by LpxT, which occurs in standard growth media, is indicated in brown. B) Inducible modifications to P. aeruginosa lipid A are indicated in color, including addition of a palmitate chain by PagP (green), removal of the 3-hydroxydecanoate acyl chain by PagL (pink), hydroxylation of the C12 secondary acyl chain(s) by LpxO (orange), addition of L-Ara4N at the lipid A phosphate groups by ArnT (blue), and pEtN addition by EptA (red).

Transcription of lipid A modification enzymes is often induced through two-component system signaling (Needham and Trent, 2013). Signal transduction occurs when a bacterial sensor kinase autophosphorylates in response to an external stimulus and transfers this phosphate group to a response regulator protein. The resulting conformation in the response regulator promotes DNA binding at target promoters to alter gene expression (Rodrigue et al., 2000). The two well-conserved two-component systems PhoPQ and PmrAB play a major role in modulating lipid A modification enzyme expression in P. aeruginosa. PhoPQ activates pagP and arnT transcription in response to limiting Mg²⁺ (McPhee et al., 2006), while PmrAB induces arnT transcription upon sensing either limiting Mg²⁺ or subinhibitory cationic antimicrobial peptide concentrations (McPhee et al., 2003; McPhee et al., 2006).

Aside from PhoPQ and PmrAB, three additional two-component systems in P. aeruginosa are involved in lipid A modification and resistance to the cationic antimicrobial peptide polymyxin. These include ParRS and CprRS, which are both activated by various antimicrobial peptides (Fernandez et al., 2010; Fernandez et al., 2012), and ColRS, which remains largely unstudied in P. aeruginosa but plays a role in heavy metal tolerance and overall membrane stability in P. putida and P. fluorescens (de Weert et al., 2006; Hu and Zhao, 2007; Ainsaar et al., 2014). A recent study investigated the role of CprRS and ColRS in P. aeruginosa polymyxin resistance. This report revealed that these two systems contribute to the elevated polymyxin resistance observed in P. aeruginosa phoQ mutants, and also suggested that ColRS and CprRS regulate additional unknown factors required for this resistance (Miller et al., 2011; Gutu et al., 2013). Given the complexities of two-component systems in P. aeruginosa, there are still significant questions regarding LPS modifications and their impact on antimicrobial resistance.

In Gram-negatives such as E. coli, Helicobacter pylori, Campylobacter jejuni, and Neisseria gonorrhoeae addition of phosphoethanolamine (pEtN) groups to lipid A by the enzyme EptA also promote polymyxin resistance and virulence (Tran et al., 2006; Cullen and Trent, 2010; Herrera et al., 2010; Cullen et al., 2011; Cullen et al., 2013; Hobbs et al., 2013). Although P. aeruginosa has pEtN transferase orthologs, pEtN addition has not been observed in P. aeruginosa lipid A prepared from cells grown under conditions that induce this modification in other organisms or from P. aeruginosa clinical isolates. We investigated whether or not P. aeruginosa lipid A could be modified with pEtN, and if so, under what conditions. Here we report that P. aeruginosa gene PA14_39020 is a functional pEtN transferase which we have named EptA_Pa that strictly modifes the 4' phosphate group of lipid A. We also demonstrate that zinc acts as a signal to induce eptA_Pa transcription via the ColRS two-component system (Fig. 7). While transcription of eptA_Pa is upregulated in response to Zn²⁺, arnT transcription is downregulated, suggesting that mechanisms are in place to mediate strict control over specific lipid A modifications. The existence of eptA_Pa reveals the potential for greater diversity in Pseudomonas lipid A structure and the versatility of the outer membrane.

Fig 7.

Proposed model of pEtN addition to P. aeruginosa lipid A. Upon sensing excess Zn²⁺, the ColS sensor kinase (green) autophosphorylates and transfers a phosphate group to the response regulator ColR (green). ColR then acts as a transcription factor, inducing transcription of eptA_Pa (red) while inhibiting that of arnT (blue). EptA_Pa protein is synthesized and transfers pEtN to the 4′-phosphate group of lipid A in the inner membrane. Lipid A is then transported to the bacterial cell surface. Following transport to the outer membrane, the 3-hydroxydecanoate acyl chain is removed by PagL (indicated in the model). In some instances, PagP can modify the lipid A (not shown). Cellular components are labelled as follows: OM, outer membrane; P, periplasm; IM, inner membrane; C, cytoplasm).

Results

P. aeruginosa has a functional EptA enzyme

In silico analysis identified three P. aeruginosa eptA orthologs with significant identity to the S. enterica eptA (pmrC) ortholog (Lee et al., 2004). These include PA14_58610 (24% identity, E-value 3e-28), PA14_21210 (43%, E-value 1e-149) and PA14_39020 (43%, E-value 6e-149). Since pEtN-modified P. aeruginosa lipid A has not been previously reported, we tested whether these orthologs could function as a lipid A pEtN transferase by expressing each gene in trans in the E. coli eptA mutant (W3110ΔeptA_Ec). ³²P-labeled lipid A was prepared and separated by TLC from these and relevant control strains including wild-type strain W3110 and W3110ΔeptA_Ec+empty vector.

While no pEtN is detected in lipid A prepared from W3110 or W3110ΔeptA_Ec + empty vector (Fig. S1A, lanes 1 and 2), expression of both eptA_Ec and PA14_39020 resulted in pEtN-modified lipid A (Fig. S1A, lanes 3 and 6). MALDI-TOF mass spectrometry (MS) analysis also confirmed that while W3110ΔeptA_Ec expressing empty vector had no pEtN (Fig. S1B), lipid A prepared from W3110ΔeptA_Ec+PA14_39020 (peptA_Pa) was modified with pEtN, as evidenced by the ion of m/z 1920.4 (Fig. S1C; predicted [M-H]- at m/z 1920.2).

Since EptA_Pa can add pEtN to E. coli lipid A, we next determined whether EptA_Pa modified P. aeruginosa lipid A. ³²P-labeled lipid A from wild-type strain PA14, PA14+empty vector or PA14+eptA_Pa was analyzed. TLC separation of lipid A clearly demonstrated an altered profile of PA14 expressing peptA_Pa (Fig. 2A, lane 3) relative to PA14 or PA14+empty vector (Fig. 2A, lanes 1 and 2). Penta-acylated, palmitoylated and L-Ara4N-modified lipid A species were observed by MALDI-TOF MS analysis of lipid A isolated from PA14+empty vector (Fig. 2B). Whereas expression of eptA_Pa in PA14 resulted in abundant ions of m/z 1489.0 and 1727.3 which correspond to the pEtN-modified lipid A species (Fig. 2C; predicted [M-H]- at m/z 1489.9 and 1728.1, respectively). These results reveal that P. aeruginosa has a functional lipid A pEtN-transferase enzyme capable of modifying both E. coli and P. aeruginosa LPS.

Fig 2.

Heterologous expression of a P. aeruginosa eptA ortholog results in pEtN addition to the lipid A. A) Cells were grown in MOPS minimal medium. Major ³²P-labeled lipid A species are indicated with a cartoon corresponding to the lipid A structure; colors of modification groups are the same as those used in Fig. 1. Expression of PA14_39020 (eptA_Pa) in P. aeruginosa results in modified lipid A species. B) MALDI-TOF MS analysis of PA14 + empty vector grown in MOPS minimal medium reveals no pEtN addition to the molecule, while C) analysis of PA14 + peptA_Pa shows pEtN modification of the lipid A. The fractions most representative of pEtN modification are shown.

EptA_Pa adds pEtN strictly to the lipid A 4' phosphate group

Previous work from our laboratory has demonstrated that in P. aeruginosa, S. enterica and E. coli the position of lipid A modification can be important due to potential competition with other modification groups. For example in S. enterica and E. coli EptA preferentially adds pEtN to the lipid A 1-phosphate group, which is the sole site of phosphorylation by the kinase LpxT (Herrera et al., 2010). Environmental conditions that activate eptA transcription simultaneously inhibit LpxT activity to prevent competition (Herrera et al., 2010; Kato et al., 2012). However, since both LpxT and ArnT enzymes in P. aeruginosa can act on either lipid A phosphate group (Bhat et al., 1990; Nowicki et al., 2014) we questioned whether EptA_Pa also has dual positional activity. To determine this we removed the 1- or the 4'-phosphate group of lipid A by heterologous expression of Fransicella novicida LpxE or LpxF phosphatases (Wang et al., 2004; Wang et al., 2006), respectively, and tested the ability of EptA_Pa to modify lipid A. This experiment was done in E. coli strain BN2 (Needham et al., 2013) since its lipid A is penta-acylated, and LpxF can only act on penta-acylated lipid A (Wang et al., 2006). BN2 also lacks some lipid A modification machinery including LpxT to facilitate easier analysis of the lipid A profiles.

Expression of eptA_Pa resulted in a pEtN-modified lipid A species that migrated below the unmodified bis-phosphorylated species (Fig. 3A, lane 2). Expression of either lpxE or lpxF caused a marked increase of mono-phosphorylated lipid A, which migrates near the top of the TLC plate (Fig. 3A, lanes 3 and 5). Simultaneous expression of eptA_Pa and lpxE resulted in the appearance of a species that migrated a distance between that of unmodified bis-phosphorylated lipid A and pEtN-modified lipid A (Fig. 3A, lane 4). MALDI-TOF MS confirmed this species to be pEtN-modified mono-phosphorylated lipid A (Fig. 3B). This same pEtN-modified species was not detected when LpxF was expressed with eptA_Pa (Fig 3A, lane 6). The lack of pEtN addition to 4'-dephosphorylated lipid A was further confirmed by MALDI-TOF MS as simultaneous expression of EptA_Pa and LpxF resulted in only a single ion of m/z 1506.1, corresponding to mono-phosphorylated lipid A (Fig. 3C; predicted [M-H]- at m/z 1507.1).

Fig 3.

EptA_Pa adds pEtN exclusively to the lipid A 4' phosphate group. A) Cells were grown in LB broth. Major ³²P-labeled lipid A species are indicated with a cartoon corresponding to the lipid A structure; colors of modification groups are the same as those used in Fig. 1. Heterologous expression of eptA_Pa in BN2 results in a pEtN-modified species, while expression of either the lpxE_Fn or lpxF_Fn phosphatase results in an increased of monophosphorylated species. Co-expression of lpxE_Fn and eptA_Pa results in pEtN addition to the 1-dephosphorylated lipid A molecule, while no pEtN addition of 4' –dephosphorylated species is detected. B) MALDI-TOF analysis of lipid A isolated from BN2 coexpressing lpxE_Fn and eptA_Pa corroborates the presence of a monophosphorylated, pEtN-modified species. C) MALDI-TOF analysis of lipid A isolated from BN2 coexpressing lpxF_Fn and eptA_Pa reveals that when the 4' phosphate group is removed, pEtN addition does not occur.

EptA_Pa-dependent addition of pEtN to the 4' phosphate group of BN2 and PA14 lipid A was corroborated by ultraviolet photodissociation (UVPD) tandem MS (Figs. S2, S3 and S4). For all UVPD mass spectra, cleavage sites (7) and (8) provided evidence of the presence of a pEtN group. In addition, the glycosidic and cross-ring cleavages at cleavage sites (10) and (11) in the fragmentation map shown in Figure S2, sites (9)-(12) in Figure S3, and sites (9)-(12) and (17)-(21) in Figure S4 further support the location of the pEtN modification at the 4' phosphate group of each lipid A species. Taken together, these results demonstrate that EptA_Pa functions strictly at the 4' phosphate group, unlike any previously characterized EptA enzyme.

Extracellular zinc induces pEtN addition to lipid A

In S. enterica, modification of lipid A with pEtN is induced via the PmrAB two-component system in response to mildly acidic pH (Perez and Groisman, 2007), and indirectly via the PhoPQ system when Mg²⁺ is limiting or cationic antimicrobial peptides are present (Kox et al., 2000; Bader et al., 2005). Since none of these signals induced pEtN addition to P. aeruginosa lipid A (data not shown), we next tested transition metals including Fe³⁺ and Zn²⁺, which activate PmrAB in E. coli (Hagiwara et al., 2004; Lee et al., 2005). We also tested Ga³⁺ due to its chemical similarity to Fe³⁺, and Cd²⁺, which is closely related to Zn²⁺ (Laddaga and Silver, 1985; Kaneko et al., 2007). Transition metals Co²⁺, Cu²⁺, Mn²⁺, and Ni²⁺ were also tested for their ability to induce pEtN modification of lipid A since they are associated with biological catalysts and are commonly found in the environment (Andreini et al., 2008; Mathiyazhagan and Natarajan, 2011).

Lipid A was isolated from ³²P-labeled P. aeruginosa grown in LB alone or supplemented with metal. Addition of extracellular Zn²⁺ but no other metal tested resulted in modified lipid A (Fig. 4A, lane 3). Zn²⁺-dependent modification was abolished when eptA_Pa was deleted from the genome and restored upon complementation with the native eptA_Pa promoter (Figs. 4A, lanes 4 and 5), suggesting that the changes observed were due to EptA_Pa activity. To determine whether Zn²⁺ induced transcription of eptA_Pa, cDNA was prepared from cells grown in the presence or absence of Zn²⁺. As shown by both quantitative and semi-quantitative reverse-transcriptase (RT) PCR, eptA_Pa gene expression is induced when Zn²⁺ is added to the growth media (Figs. 4B and S5). These results indicate that transcription of eptA_Pa is dependent on extracellular Zn²⁺

Fig 4.

Zn²⁺ induces transcription of eptA_Pa. A) Cells were grown in LB broth. Major ³²P-labeled lipid A species are indicated with a cartoon corresponding to the lipid A structure; colors of modification groups are the same as those used in Fig. 1. Both heterologous expression of eptA_Pa as well as addition of 2mM ZnSO₄ to the media results in pEtN addition to lipid A. This modification is not detectable in the eptA_Pa mutant, but restored upon complementation with peptA_nprom. B) Relative gene expression of eptA_Pa and arnT in response to Zn²⁺. Transcription of eptA_Pa is induced by 2mM ZnSO₄ approximately 21-fold. Zn²⁺ downregulates arnT transcription by >4-fold. Ratios were standardized relative to expression of the housekeeping control gene, clpX. C), D) and E). MALDI-TOF MS analysis of lipid A prepared from cells grown in LB broth. C) Analysis of PA14 + 2mM ZnSO₄ reveals pEtN addition to lipid A, while D) ΔeptA_Pa + 2mM ZnSO₄ shows no pEtN modification, but instead L-Ara4N addition. E) Complementation of ΔeptA_Pa with peptA_nprom restores pEtN addition to the lipid A in response to Zn²⁺. The fractions most representative of pEtN modification are shown.

To confirm whether Zn²⁺-dependent induction of EptA_Pa synthesis resulted in pEtN addition to P. aeruginosa lipid A, MALDI-TOF MS analysis of lipid A isolated from cells grown with or without Zn²⁺ was performed. Wild-type P. aeruginosa lipid A was modified with pEtN in the presence of Zn²⁺ (Fig. 4C), while lipid A prepared from the PA14ΔeptA_Pa mutant showed no pEtN modification when Zn²⁺ was added to the media (Fig. 4D). Complementation of PA14ΔeptA_Pa using the native eptA_Pa promoter restored Zn²⁺-dependent pEtN addition to the lipid A (Fig. 4E). MS analysis revealed that in addition to pEtN addition, L-Ara4N-modified lipid A was present in PA14 grown with Zn²⁺ (Fig. 4C). We were therefore curious as to how Zn²⁺ might influence L-Ara4N addition to lipid A. Since Zn²⁺ affected the transcription of eptA_Pa, we assessed whether Zn²⁺ altered arnT gene expression by performing quantitative RT-PCR. While eptA_Pa transcription increased by 21-fold in the presence of 2mM Zn²⁺, arnT transcription was downregulated >4-fold (Fig. 4B). This result indicates that pEtN modification is selected for in the presence of Zn²⁺ while arnT expression, and thus L-Ara4N modification of lipid A, is downregulated.

The ColRS two-component system induces pEtN addition to lipid A

Since inducible lipid A modification genes like eptA are commonly regulated by two-component systems, our next goal was to determine the system responsible for eptA_Pa transcriptional activation in response to Zn²⁺. The ColRS system has recently been shown to respond to transition metals including Zn²⁺ (Ainsaar et al., 2014). Our first approach was therefore to investigate the presence of potential ColR binding sites in the eptA_Pa promoter. A consensus ColR binding site has been determined for promoters of genes within the ColR regulon in P. putida (Kivistik et al., 2009). Using the Virtual Footprint online analysis tool (Münch et al., 2005), we found three potential ColR binding sites within the eptA_Pa promoter region with close agreement to this consensus sequence (Fig 5A), suggesting that ColR binds to the eptA_Pa promoter. It was then tested whether overexpression of colR could induce eptA_Pa transcription by semi-quantitative RT-PCR of cDNA. Response regulators pmrA and phoP, which directly or indirectly regulate eptA transcription in S. enterica, were also tested. Only overexpression of colR resulted in detectable transcription of eptA_Pa (Fig 5B). Lipid A was modified with pEtN upon overexpression of colR, as demonstrated by both TLC separation of ³²Plabeled lipid A (Fig 5C, lane 4) and MALDI-TOF MS analysis of PA14 + pcolR (Fig 5D).

Fig 5.

The two-component system response regulator ColR activates eptA_Pa transcription. A) Putative eptA_Pa promoter ColR binding sites are in bold and boxed; nucleotides that deviate from the conserved recognition sequence in P. putida ((T/C)(T/C)NA(C/G)NN(T/C)TTTTT(C/G)AC) are indicated in red. The number of base pairs between ColR sites or upstream of the start codon is indicated. B) Semi-quantitative RT-PCR of cDNA prepared from cells grown in MOPS minimal medium. While eptA_Pa is not transcribed in PA14, expression of colR in trans results in eptA_Pa transcription. C) Lipid A was isolated from ³²P-labeled cells grown in MOPS minimal medium and separated by TLC. Only expression of the colR response regulator, and not pmrA or phoP, results in pEtN modification of lipid A. D) MALDI-TOF MS analysis of PA14 + pcolR grown in MOPS minimal medium reveals pEtN-modified lipid A. The fraction most representative of pEtN modification is shown.

To determine whether ColR induction of eptA_Pa transcription is dependent on Zn²⁺, a PA14 colR deletion mutant was generated and assessed for transcription of eptA_Pa in the presence or absence of Zn²⁺ by both quantitative and semi-quantitative RT-PCR analysis. Although a Zn²⁺ concentration of 2mM had been used in the initial Zn²⁺ assay experiments, the PA14ΔcolR mutant was sensitive to 2mM Zn²⁺. Instead, 1mM Zn²⁺ was used, which was sufficient to visualize pEtN modification in PA14 (Fig. 6A, lane 2). Minimal eptA_Pa transcription was detected in response to Zn²⁺ upon deletion of colR; complementation of this mutant restored Zn²⁺-dependent eptA_Pa transcription by >4-fold (Fig. 6B, Fig. S5). While 1mM Zn²⁺ induced pEtN modification of PA14 lipid A (Fig. 6A, lane 2), lipid A from the PA14ΔcolR was not modified with pEtN in response to Zn²⁺, as determined by both TLC separation of ³²P-labeled lipid A and MALDI-TOF MS analysis (Fig. 6A, lane 3 and Fig 6C). pEtN addition was restored upon complementation of the colR mutant with pcolR_nprom (Fig. 6A, lane 4 and Fig 6D). A PA14 colS mutant and complemented mutant were also tested for pEtN addition to lipid A in response to Zn²⁺ by TLC separation of ³²P-labeled lipid A. As for the colR mutant, lipid A modification with pEtN was not detected in the colS mutant grown in the presence of 1mM Zn²⁺, but was restored upon complementation of colS with pcolS_nprom (Fig. S6, lanes 3 and 4). These results demonstrate that the ColRS system induces pEtN addition to lipid A upon sensing Zn²⁺.

Fig 6.

Deletion of colR results in loss of Zn²⁺-induced pEtN modification of P. aeruginosa lipid A. A) Lipid A was isolated from ³²P-labeled cells grown in LB broth and separated by TLC. While pEtN modification of lipid A is detectable for PA14 + 1mM ZnSO₄, no such modification occurs in PA14ΔcolR in response to Zn²⁺. Modification is restored in the complemented mutant. B) Relative gene expression of eptA_Pa and arnT in response to Zn²⁺ in the ΔcolR mutant or complemented mutant. Transcription of eptA_Pa in the presence of 1mM ZnSO₄ is induced >4-fold in a ColR-dependent manner. An approximately 10-fold decrease in arnT transcription in the presence of 1mM ZnSO₄ is also dependent on ColR. Ratios were standardized relative to expression of the housekeeping control gene, clpX. C) and D). MALDI-TOF MS analysis of lipid A prepared from cells grown in LB broth. C) No pEtN modification is detected in the PA14ΔcolR mutant grown in LB + 1mM ZnSO₄. D) Complementation of PA14ΔcolR with pcolR_nprom restores the Zn²⁺-dependent pEtN addition to the lipid A. The fractions most representative of pEtN modification are shown.

We also investigated whether the downregulation of arnT transcription in the presence of Zn²⁺ was dependent on the ColR response regulator. Gene expression of arnT was analyzed by quantitative RT-PCR in the PA14 colR mutant and complemented strains in the presence of 1mM Zn²⁺. Transcription of arnT was reduced by approximately 10-fold upon complementation of colR (Fig. 6B). This result indicates that ColR activates transcription of eptA_Pa in response to Zn²⁺ while downregulating arnT transcription.

Discussion

Changes in the environment require bacterial outer membrane remodeling, including LPS structural changes, to promote membrane stability (Whitfield and Trent, 2014). L-Ara4N addition to lipid A phosphate groups contributes to cationic antimicrobial peptide resistance in P. aeruginosa, E. coli and S. enterica (Lee et al., 2004; Herrera et al., 2010; Fernández et al., 2013). The addition of the amine-containing residue pEtN can also result in increased peptide resistance (Tran et al., 2006; Herrera et al., 2010), and in some organisms is a crucial factor for host infection (Cullen et al., 2013; Hobbs et al., 2013). Lipid A modifications in P. aeruginosa have been well-studied, yet despite the existence of eptA orthologs, pEtN addition has never been observed. Due to the importance of pEtN lipid A modification in other organisms, we investigated the functionality and regulation of P. aeruginosa eptA orthologs. In this report, we identify and characterize a functional P. aeruginosa lipid A pEtN transferase and determine that Zn²⁺ induces transcription of eptA_Pa via the Pseudomonas-specific ColRS system (Fig. 7).

Overexpression of three P. aeruginosa eptA orthologs in E. coli revealed that PA14_39020 (eptA_Pa) was able to modify lipid A with pEtN (Fig. S1A and C). It is likely that the other two orthologs add pEtN to other targets in the cell. Based on its homology to S. enterica CptA, PA14_58610 may be the enzyme responsible for adding pEtN to the core of P. aeruginosa LPS (Kooistra et al., 2003; Tamayo et al., 2005). While a very minor amount of lipid A modification is detected by TLC separation of lipid A upon expression of PA14_21210 in PA14, lipid A is probably not the primary target of this enzyme. It is possible that this enzyme modifies an as yet unidentified target, as pEtN transferase enzymes in other organisms have been shown to modify structural proteins of the flagellum and pilus (Hegge et al., 2004; Cullen and Trent, 2010), and in doing so has some very minor, non-specific activity toward lipid A. This activity toward lipid A, however, is so minor that is cannot be detected by mass spectrometry analysis (data not shown).

We characterized the site-specificity of pEtN addition to lipid A due to the potential for competition with other modification groups. Whereas pEtN addition occurs specifically or preferentially at the 1-phosphate group of lipid A in H. pylori (Tran et al., 2004) and S. enterica (Herrera et al., 2010), respectively, analysis of EptA_Pa activity in E. coli revealed that this enzyme acts solely at the 4' position (Figs. 3, S2). EptA_Pa activity thus differs from ArnT and LpxT enzymes in P. aeruginosa that can modify either lipid A phosphate group (Bhat et al., 1990; Nowicki et al., 2014).

Investigation of conditions that induce pEtN modification revealed that excess Zn²⁺ acts as the activating signal for eptA_Pa transcription. Pseudomonas species are readily found in the soil and aqueous environments which can be contaminated with metals due to waste runoff from mines, smelting, and other industrial facilities (Teitzel and Parsek, 2003; Raja et al., 2006; Mathiyazhagan and Natarajan, 2011). In such environments, Pseudomonas can be exposed to high levels of metal pollutants and has thus evolved the ability to alter gene expression to promote metal tolerance (Perron et al., 2004; Ha et al., 2004; Hu and Zhao, 2007; Caille et al., 2007). Excess Zn²⁺ may also be relevant in healthcare settings as concentrations up to 1mM can leach out from latex catheters and gloves (Perron et al., 2004; Ballesta et al., 2006). Deletion of eptA_Pa, however, does not result in increased sensitivity to Zn²⁺ or to Cd²⁺, Ga³⁺, Fe³⁺, Co²⁺, Cu²⁺, Mn²⁺ or Ni²⁺. Under laboratory settings Zn²⁺-induced pEtN addition to lipid A has no effect on polymyxin resistance, biocide tolerance, or biofilm formation (data not shown). As the ColRS system was previously implicated to play a role in polymyxin resistance, unidentified genes other than eptA_Pa within the ColRS regulon are likely involved in this resistance. The fact that P. aeruginosa has evolved regulatory mechanisms to control pEtN addition to lipid A, however, suggests the importance of this modification for conditions we have not yet identified.

Extracellular metals are sensed by one of three two-component systems in Pseudomonas species: CzcRS (Perron et al., 2004), CopRS (Caille et al., 2007) and ColRS (Hu and Zhao, 2007; Ainsaar et al., 2014). Both CzcRS (activated by Zn²⁺) and CopRS (activated by Cu²⁺) induce expression of the heavy metal efflux pump CzcCBA while downregulating the OprD porin, leading to decreased carbapenem and imipenem uptake (Perron et al., 2004; Caille et al., 2007). In P. putida, the ColRS system senses Zn²⁺, Fe³⁺, Mn²⁺ and Cd²⁺, and mutants of colR and colS display lower tolerance to these metals (Ainsaar et al., 2014). We have now determined Zn²⁺ to be an activating signal for ColRS in P. aeruginosa. Additionally, the PA14 colR and colS mutants are more sensitive to Zn²⁺ than wild-type, suggesting a role for the ColRS system in Zn²⁺ tolerance in P. aeruginosa. While the ColRS system is important for metal tolerance in P. putida and deletion of multiple genes in the ColRS regulon results in increased metal sensitivity, no individual ColRS-regulated gene has a major contribution to metal tolerance (Ainsaar et al., 2014). It is likely that multiple genes in the P. aeruginosa ColRS regulon are involved in Zn²⁺ tolerance, which could explain why the eptA_Pa isogenic mutant does not have any growth defect in media with Zn²⁺.

Our demonstration of pEtN-modified lipid A via the ColRS system in response to Zn²⁺ (Fig. 7) reveals that lipid A remodeling in P. aeruginosa is more complex than previously realized. Extracellular Zn²⁺ specifically induces transcription of eptA_Pa and not arnT through the ColR response regulator, demonstrating coordinated control over lipid A modifications. In addition to selectively inducing expression of eptA_Pa Zn²⁺ downregulates arnT transcription by over 4-fold (Fig. 4B). This is interesting given that ArnT-mediated L-Ara4N modification typically plays a more significant role in virulence and antimicrobial peptide resistance than eptA in organisms possessing both modification enzymes (Tamayo et al., 2005; Herrera et al., 2010). While eptA_Pa does not seem to be involved in metal or polymyxin resistance, there is likely an evolutionary reason for this targeted induction of eptA_Pa transcription. Our findings demonstrate the tight control of P. aeruginosa lipid A modification systems, and suggest the need for further studies to better elucidate the mechanisms involved in outer membrane remodeling and its contribution to bacterial persistence and versatility.

Experimental Procedures

Bacterial strains and growth conditions

The bacterial strains and plasmids used in this study are listed in Table S1. E. coli strains were cultured in LB broth or agar (Difco) at 37°C. P. aeruginosa strains were grown on LB agar plates, and initial liquid cultures were grown overnight in LB broth at 37°C. The next day, P. aeruginosa cultures were diluted to an OD₆₀₀ of ~0.05 in either LB broth or in morpholinepropanesulfonic acid (MOPS)-buffered minimal medium (50mM MOPS, 93mM NH₄Cl, 43mM NaCl, 2mM KH₂PO₄) supplemented with 3.5μM FeSO₄•7H₂O, 20mM sodium succinate, and 1mM MgSO₄. Chloramphenicol was used at a concentration of 30μg/mL for E. coli. Ampicillin or carbenicillin was used at a concentration of 100μg/mL or 300μg/mL for E. coli or P. aeruginosa, respectively. For growth of P. aeruginosa in medium with added metals, LB was used to prevent metals from crashing out of solution. For the initial screen of lipid A modifications, metal salts were added in the following concentrations: 2mM ZnSO₄, 0.2mM CdSO₄, 0.1mM Ga(III)NO₃, 0.2mM FeSO₄, 0.1mM CoCl₂, 2mM CuSO₄, 2mM MnSO₄, and 2mM NiSO₄. The highest concentration of metal that did not significantly reduce growth (defined as a >50% reduction) in liquid medium was used, with the exception of Fe³⁺, for which a lower, more physiologically relevant concentration was used based on concentrations known to induce lipid A modification in other organisms (Herrera et al., 2010).

DNA and RNA preparation

Before preparing P. aeruginosa genomic DNA from an overnight culture in LB broth, two washes with 0.1M NaCl were performed. Genomic DNA was prepared using the Easy-DNA Kit (Invitrogen). Total RNA was extracted from cells grown to an OD₆₀₀ of ~0.6 using the RNeasy Mini Kit (Qiagen), according to the manufacturer's instructions. To eliminate residual DNA contamination, total RNA was treated with DNase from the RNase-Free DNase Set (Qiagen). cDNA synthesis was performed using SuperScript III Reverse Transcriptase (Invitrogen) according to the manufacturer's instructions.

Recombinant DNA methods

Plasmid DNA was isolated with the QIAprep Spin Miniprep Kit (Qiagen). Chromosomal DNA for insertion into plasmid constructs was amplified using either the DNA polymerase PfuTurbo^R (Stratagene) or Takara Ex Taq (Takara). PCR products were separated on an agarose gel and purified using the QIAquick Gel Extraction Kit (Qiagen). All primers were purchased from Integrated DNA Technologies (Table S2). Restriction endonucleases, T4 DNA ligase, and Antarctic Phosphatase used in this study for generation of plasmid constructs were purchased from New England Biolabs and used according to the manufacturer's instructions.

Generation of chromosomal gene deletion mutants

In-frame, markerless gene deletions were generated in P. aeruginosa by homologous recombination using the suicide plasmid pEX18Gm. ~1Kb DNA fragments flanking the target gene up or downstream were amplified using primers listed in Table S2. An assembly PCR was then carried out to stitch together these flanking regions. Assembly PCR fragments were digested with restriction endonucleases EcoRI and HindIII or BamHI, and ligated into pEX18Gm. The suicide plasmid constructs, pEX18-eptAdel, pEX18-colRdel or pEX18-colSdel were introduced into P. aeruginosa via conjugation with E. coli strain SM10. Deletion mutants were then screened for as described previously (Hoang et al., 1998). Deletions were confirmed by PCR.

Plasmid constructs

To construct pPA14_39020 (pACeptA_Ec), pPA14_58610 and pPA14_21210, each gene was amplified along with the native RBS and cloned into the medium copy vector pACYC184 using EcoRV and SalI restriction endonucleases. For generation of peptA_Pa, pcolR, ppmrA and pphoP, each gene and its native RBS were amplified and digested with EcoRI and HindIII to clone into pEX1.8. All constructs were confirmed by sequencing. The eptA_Pa and colR genes were amplified along with their native promoters and cloned into the medium copy vector pEX1.8 (Pearson et al., 1997) by digestion with SalI or SalI and HindIII, respectively, generating peptA_nprom and p colR_nprom. For generation of pcolS_nprom, the colRS promoter was first amplified and cloned into pEX1.8 by digestion with BamHI and EcoRI. The colS coding sequence was then amplified and cloned into pEX1.8 (containing the colRS promoter) with EcoRI and HindIII.

Isolation and analysis of labeled lipid A

Overnight cultures were diluted to an OD₆₀₀ of ~0.05 in 5mL of fresh medium (as indicated within each figure legend) and labeled with 2.5μCi/mL ³²P_i (Perkin-Elmer). Cells were harvested at an OD₆₀₀ of ~1.0, and lipid A was isolated by mild acid hydrolysis as described previously (Zhou et al., 1999; Tran et al., 2004). ³²P_i-labeled lipid A species were spotted on a TLC plate at ~5,000 cpm per lane (10,000 cpm for E. coli), and run in a solvent prepared in a 50:50:16:5 (v/v) ratio of chloroform, pyridine, 88% formic acid, and water, respectively. TLC plates were dried, set on a phosphor screen overnight and imaged using a phosphor-imager (BioRad PMI).

Large scale lipid A isolation and MALDI-TOF mass spectrometry

For large scale lipid A analysis, 250mL cultures were grown at 37°C to an OD₆₀₀ of ~1.0 in the medium indicated. Lipid A was prepared by mild acid hydrolysis, washed and resuspended in chloroform/methanol/water (2:3:1, v/v), as described previously (Hankins et al., 2013). The sample was run through a DEAE cellulose column, washed with chloroform/methanol/water (2:3:1, v/v), and eluted in individual fractions with chloroform/methanol/increasing concentrations of ammonium acetate, as described previously (Odegaard et al., 1997; Hankins et al., 2013). Typically, hydrophilic or monophosphorylated species fractionate at lower ammonium acetate concentrations (flow-through, wash, 60 or 120mM elution fractions), while more hydrophobic, unmodified or phosphate-modified species fractionate at the highest concentration, 480mM ammonium acetate. MALDI-TOF mass spectrometry was performed as described using a MALDI-TOF/TOF mass spectrometer (ABI 4700 Proteomics Analyser) (Hankins et al., 2013).

ESI and UVPD mass spectrometry

Lipid A was isolated and prepared as described above for MALDI-TOF analysis. All mass spectrometry experiments were executed on a Thermo Scientific Orbitrap Elite mass spectrometer (Bremen, Germany) modified to perform ultraviolet photodissociation (UVPD). The mass spectrometer was equipped with a 193-nm Coherent ExciStar XS excimer laser (Santa Clara, CA) and operated in the negative ion mode using a previously described set-up (Shaw et al., 2013). Briefly, solutions containing 1–5 μM lipid A in 50:50 methanol/ chloroform were directly infused using an electrospray ionization (ESI) source at a flow rate of 3 μl/min. The ESI voltage was set to 4 kV. UVPD mass spectra were collected using 10 laser pulses per spectrum (at 4–5 mJ/pulse) and were interpreted as described previously (Madsen et al., 2011).

Quantitative PCR methods

Primers for semi-quantitative and quantitative PCR (qPCR) were designed using the Primer-BLAST tool (NCBI) and are listed in Table S2. Semi-quantitative PCR was performed by amplifying cDNA obtained from samples cultured in the conditions or with the ppmrA, pphoP and pcolR expression constructs as indicated, using primers specific for eptA_Pa or clpX as a reference gene (Palmer et al., 2005). qPCR was performed in a OneStep thermocycler (Applied Biosystems) using Power SYBR Green PCR Master Mix (Applied Biosystems), according to the manufacturer's instructions, as described previously (Pfaffl, 2001; Nowicki et al., 2014).