Structures of two bacterial resistance factors mediating tRNA-dependent aminoacylation of phosphatidylglycerol with lysine or alanine

Significance

Lipid homeostasis is a fundamental process for understanding antimicrobial susceptibility. Modification of the polar head group of phosphatidylglycerol into the respective aminoacyl-ester of phosphatidylglycerol is a widely used strategy to mediate bacterial resistance. Here we present the structures of the catalytic domains of aminoacyl-phosphatidylglycerol synthases from Pseudomonas aeruginosa and Bacillus licheniformis. These prototypical enzymes specifically catalyze the tRNA-dependent synthesis of alanyl-phosphatidylglycerol and lysyl-phosphatidylglycerol, respectively. A central tunnel architecture facilitates binding of the polar aminoacyl-tRNA molecule opposite the hydrophobic lipid substrate as a fundamental principle for the catalysis at the water–lipid interface. Specific inhibition of aminoacyl-phosphatidylglycerol synthases might be a promising strategy to render Gram-positive and Gram-negative pathogenic bacteria more susceptible to antimicrobial treatment.

Abstract

The cytoplasmic membrane is probably the most important physical barrier between microbes and the surrounding habitat. Aminoacylation of the polar head group of the phospholipid phosphatidylglycerol (PG) catalyzed by Ala-tRNA^Ala–dependent alanyl-phosphatidylglycerol synthase (A-PGS) or by Lys-tRNA^Lys–dependent lysyl-phosphatidylglycerol synthase (L-PGS) enables bacteria to cope with cationic peptides that are harmful to the integrity of the cell membrane. Accordingly, these synthases also have been designated as multiple peptide resistance factors (MprF). They consist of a separable C-terminal catalytic domain and an N-terminal transmembrane flippase domain. Here we present the X-ray crystallographic structure of the catalytic domain of A-PGS from the opportunistic human pathogen Pseudomonas aeruginosa. In parallel, the structure of the related lysyl-phosphatidylglycerol–specific L-PGS domain from Bacillus licheniformis in complex with the substrate analog L-lysine amide is presented. Both proteins reveal a continuous tunnel that allows the hydrophobic lipid substrate PG and the polar aminoacyl-tRNA substrate to access the catalytic site from opposite directions. Substrate recognition of A-PGS versus L-PGS was investigated using misacylated tRNA variants. The structural work presented here in combination with biochemical experiments using artificial tRNA or artificial lipid substrates reveals the tRNA acceptor stem, the aminoacyl moiety, and the polar head group of PG as the main determinants for substrate recognition. A mutagenesis approach yielded the complementary amino acid determinants of tRNA interaction. These results have broad implications for the design of L-PGS and A-PGS inhibitors that could render microbial pathogens more susceptible to antimicrobial compounds.

Bacteria can adapt rapidly to changing environmental conditions by controlling the physical properties of biological membranes (1). One important strategy is the tRNA-dependent aminoacylation of the polar head group of phosphatidylglycerol (PG) by aminoacyl-phosphatidylglycerol synthases (aa-PGSs). The resulting products alanyl-phosphatidylglycerol (A-PG) or lysyl-phosphatidylglycerol (L-PG) (Fig. 1) reduce the overall net negative charge of the membrane, making it less susceptible to cationic antimicrobial peptides (CAMPs). Such CAMPs often are synthesized as innate immunity host-defense peptides in response to bacterial infections (2–5). Fundamental work with the Gram-positive pathogen Staphylococcus aureus indicated aa-PGS–mediated nonsusceptibility to vancomycin (an antibiotic of last resort), resistance to host antimicrobial peptides (defensins), and protection against neutrophil killing (6–9). Accordingly, aa-PGS enzymes can be considered virulence factors and thus have been termed “multiple peptide resistance factors.” In Pseudomonas aeruginosa, aminoacylation of PG also was found in response to acidic environmental conditions (2, 5), inter alia. P. aeruginosa is the dominant pathogen that colonizes the lung of patients suffering from cystic fibrosis. Because of the underlying defect in bicarbonate ion transport, acidification of the airway surface liquid contributes to cystic fibrosis pathogenesis (10). Furthermore, during inflammatory response local acidification by the production of acids was observed (11).

Fig. 1.

tRNA-dependent aminoacylation of phosphatidylglycerol with alanine or lysine. A-PGS and L-PGS (also named “MprF”) catalyzed the formation of A-PG (A) and L-PG (B), respectively.

Aa-PGS catalysis is one of the rare instances in which a ribosomal aminoacyl-tRNA is used other than in protein biosynthesis. Aa-PGS enzymes are classified as (i) A-PGS, found, for example, in Gram-negative P. aeruginosa; (ii) L-PGS, found in Gram-positive S. aureus, Bacillus anthracis, and Bacillus licheniformis, inter alia; and (iii) aa-PGS enzymes with a broadened substrate specificity (synthesis mainly of L-PG together with A-PG) as described for Bacillus subtilis and Enterococcus faecium (2–5, 12). Recent results demonstrate that precise tuning of cellular A-PG and/or L-PG concentrations is fundamental for bacterial resistance (2). Furthermore, regulatory circuits including specific aminoacyl-phosphatidylglycerol (aa-PG) hydrolases have been described for P. aeruginosa and E. faecium (13, 14).

The aa-PGS enzymes consist of a separable, water-soluble C-terminal domain showing full enzymatic activity (4, 15). This domain contains all elements for the specific recognition of the ∼26 kDa water-soluble aminoacyl-tRNA (Ala-tRNA^Ala or Lys-tRNA^Lys) and for binding of the hydrophobic PG substrate (15). The N-terminal transmembrane domain anchors aa-PGS proteins in the cytoplasmic membrane and harbors an additional translocase activity responsible for flipping the newly synthesized lipid into the outer leaflet of the cytoplasmic membrane (16).

This study presents the first, to our knowledge, structures of the catalytic A-PGS domain from P. aeruginosa together with the related L-PGS structure from B. licheniformis in the presence of a small substrate analog. Structural biology in combination with biochemical experiments using a series of artificial aminoacyl-tRNA or artificial lipid substrates allows the molecular understanding of aa-PGS substrate recognition as a basis for the future exploitation of this class of previously unidentified antimicrobial targets.

Results and Discussion

The determination of the crystal structures of the catalytic A-PGS and L-PGS domains from P. aeruginosa and B. licheniformis is summarized in Table S1 (effective resolution, 2.4 Å for each). The structural superposition of both structures with an rmsd of 1.1 Å is depicted in Fig. 2A (34.1% amino acid sequence identity; compare with alignment in Fig. S1). Despite this high degree of structural conservation, a marked specificity for the synthesis of A-PG or L-PG was observed for A-PGS and L-PGS in enzymatic assays using Ala-tRNA^Ala or Lys-tRNA^Lys as substrates, respectively (Fig. S2).

Table S1.

Data collection and refinement statistics

Dataset	A-PGS SeMet derivative	A-PGS native	L-PGS
Temperature, K	100	100	100
Wavelength, Å	0.9797	0.9184	1.0332
Δϕ, °	0.7	0.5	0.2
Space group	P4122	P4122	P1
Unit cell parameters	a = 94.0 Å	a = 94.0 Å	a = 46.9 Å
	b = 94.0 Å	b = 94.0 Å	b = 66.3 Å
	c = 166.5 Å	c = 166.5 Å	c = 71.2 Å
	α = γ = 90°	α = γ = 90°	α = 112.4°
	β = 90°	β = 90°	β = 94.1°
	γ = 90°	γ = 90°	γ = 98.9°
Resolution range dmax-dmin, Å	50.0–3.1 (3.2–3.1)	20.0–2.3 (2.4–2.3)	40–2.1 (2.2–2.1)
No. of observed reflections	315,054 (23,659)	262,982 (26,038)	128,799 (2,501)
Completeness	0.999 (0.999)	0.995 (0.991)	0.806 (0.191)
Multiplicity	12.3 (12.2)	7.8 (7.9)	3.5 (3.5)
Rmerge*	0.307 (1.557)	0.159 (1.343)	0.095 (0.596)
Average I/σ,I	12.9 (2.1)	12.3 (2.0)	8.7 (2.0)
CC(1/2)†	0.994 (0.853)	0.997 (0.729)	0.994 (0.720)
Effective dmin‡, Å	3.1	2.4	2.4
Wilson B, Å2	56.1	37.4	27.5
Solvent content	0.74	0.74	0.52
Refinement
Rwork§	0.182	0.175	0.197
Rfree¶	0.213	0.199	0.248
No. of nonhydrogen atoms per asymmetric unit
Total	2,598	2,728	5,468
Protein	2,585	2,557	5,234
No. of molecules per asymmetric unit
Protein	1	1	2
Water	—	111	192
Ligands	5	16	3
Atomic displacement factors B, Å2
Overall	50.6	48.7	36.0
Protein	50.4	48.4	36.0
Water	—	46.2	34.4
Ligands	99.8	69.4	36.5
Rmsd from ideal
Bond lengths, Å	0.002	0.008	0.002
Bond angles, °	0.64	1.11	0.54
Ramachandran plot
Favored, %	96.3	95.6	95.7
Allowed, %	2.5	3.8	3.0

Values in parentheses are for the highest-resolution shell.

^*R_merge = Σ_hkl,i(Σ_i|I_hkl,i-〈I_hkl〉|/Σ_hkl,i〈I_hkl〉.

^†Fraction of correlation between intensities from random half-datasets (48).

^‡Resolution beyond which data completeness falls below the arbitrary threshold of 0.9 or the I/σ(I) average falls below 2.

^§R_work = (Σ||F_o|-|F_c||/Σ|F_o|); |F_o|, structure factor magnitudes observed; |F_c|, structure factors calculated.

^¶R_free is computed as R_work but using 5% randomly assigned reflections excluded from refinement.

Fig. 2.

X-ray structures of the catalytic domain of A-PGS from P. aeruginosa and L-PGS from B. licheniformis. (A) Structural superposition of A-PGS (light blue, blue, purple, and gray) and L-PGS (light gray) in a wall-eye stereo view. (B) Topology diagram of the tandem repeated GNAT fold of A-PGS and L-PGS. GNAT domains 1 and 2 (light blue and blue) share strands E and K (purple). The detailed sequence assignment for A-PGS and L-PGS is depicted in Fig. S1. (C, Center) A longitudinal section of the A-PGS structure from P. aeruginosa. (Left and Right) The resulting halves reveal a tunnel which contains the upper tRNA-binding pocket and the lower PG-binding tunnel. The key catalytic residues of A-PGS (yellow sticks) and residues located at the constriction of the tunnel (light gray sticks) and the C terminus (semitransparent gray surface) are located above the respective sectional plane. The protein is represented as a van der Waals surface and is colored according to its surface electrostatics [−20 (red) to +20 (blue) K_bT/e_c]. (D, Center) Longitudinal section of L-PGS in complex with the competitive inhibitor L-lysine amide. (Left and Right) The resulting halves reveal a tunnel that contains the L-lysine amide (LYN, green ball-and-stick representation)-binding site in the upper tRNA-binding pocket (Left) and the lower PG-binding cavity. Important amino acid residues are highlighted and colored as in C. The enlarged view in the left panel shows the polar interactions of key catalytic residues (yellow sticks) with the L-lysine amide, which is an analog of the aminoacyl moiety of the substrate Lys-tRNA^Lys. The position of the structurally related A-PGS amino acid residues is superimposed (light gray sticks). (E) Superposition of the FemX/tRNA structure (20) with the L-PGS/L-lysine amide complex reveals a plausible binding mode of the aminoacylated tRNA acceptor arm. A structural overlay of L-PGS/L-lysine amide with the FemX/tRNA complex localizes the terminal tRNA nucleotides (CCA) in a defined cavity of L-PGS. The 3′ hydroxyl oxygen atom of the terminal ribose is placed only 1.0 Å away from the cocrystallized L-lysine amide (LYN, offset indicated by asterisk). (F) Model for the interaction of A-PGS and tRNA. A representative tRNA molecule (PDB ID code 1TN1) was positioned onto the structure of A-PGS. The aminoacyl-binding pocket and the CCA-binding pocket were used as reference points. Mutational analysis for amino acid residues Lys676, Arg684, Arg687, and Asn683 located on helix 5 resulted in impaired A-PGS activity. The terminal base pairings G2-C71, G3-U70, and G4-C69 have been identified previously as tRNA-recognition elements (15). The theoretical position of the U70 base is indicated by an asterisk. (G) L-lysine amide (LYN) bound within the aminoacyl-binding pocket of L-PGS (yellow bonds). Hydrogen bonds as shown as dashed lines with distances indicated in Ångstroms. Generated with LIGPLOT (43). (H) Wall-eye stereo view of the lipid-binding tunnel of A-PGS. Enlarged view of the longitudinal section of Fig. 2C, Right indicating conserved amino acid residues (gray and blue). The functional role of amino acid residues Gln636, Glu658, and Ser763 (gray) was demonstrated by mutagenesis (see also Table S2). Docking calculations [using Autodock Vina (41)] revealed the theoretical binding mode of the PG lipid substrate (green sticks, representing PG C5:0/C8:0).

Fig. S1.

Structure-based sequence alignment of the catalytic domain of A-PGS from P. aeruginosa, L-PGS from B. licheniformis, and FemX from Weissela viridescens (20, 21). Alignments were calculated using University of California, San Francisco Chimera (38, 39). Clustal Omega (44) was used for primary sequence analysis. Conserved and partly conserved residues are indicated by an asterisk (*) and colon (:) or period (.), respectively, and are named Cons1 for conservation between A-PGS sequences, Cons2 for L-PGS sequences, and Cons3 for the overall alignment of A-PGS, L-PGS, and FemX. Helices (boxes) and β-strands (arrows) are marked. GNAT domain 1 is colored in light blue, GNAT domain 2 in blue, secondary structure elements of A-PGS and L-PGS that are part of both GNAT domains are highlighted in purple, and inserted elements are colored in gray. Underlined fonts indicate L-PGS residues that are involved in L-lysine amide interaction. Green fonts indicate conserved residues of the lipid tunnel. Red fonts mark residues of FemX (and A-PGS and L-PGS), which are involved (might be involved) in the interaction with the CCA end of tRNA. Orange fonts highlight residues proposed for tRNA acceptor stem interaction. Blue fonts indicate FemX residues that are involved in pentapeptide interaction (20).

Fig. S2.

Incorporation of radioactively labeled ¹⁴C-lysine and ¹⁴C-alanine into the organic lipid phase by the enzymatic activities of the catalytic domain of A-PGS and L-PGS. The crude cellular extract of E. coli cells overproducing A-PGS_543–881 and L-PGS_519–850, respectively, were supplemented with radioactively labeled ¹⁴C-lysine (A) or ¹⁴C-alanine (B) and an ATP regenerating system as described (5). Samples were subjected to lipid extraction, and synthesis of radioactively labeled lipids was analyzed by liquid scintillation counting.

Both aa-PGS structures share a tandem repeated GNAT (GCN5-related N-acetyltransferase) fold (17). The topology diagram in Fig. 2B highlights GNAT domains 1 and 2 (colored light blue and blue, respectively) partially arranged with internal “pseudo-twofold symmetry” (rhombus), which share strands E and K (purple), respectively. Internal superposition of the central elements of GNAT domains 1 and 2 resulted in rmsds of 0.99 Å for A-PGS and 0.96 Å for L-PGS. Domain 2 shares the specific sequential arrangement of the GNAT fold (17, 18) with an insertion of two α-helices (8a and 9), whereas domain 1 lacks the most N-terminal β-strand of the superfamily.

The strongest structural homology (19) of both aa-PGS structures was found to the alanyl transferase FemX [Protein Data Bank (PDB) code 4II9], which is another Ala-tRNA^Ala–dependent enzyme involved in peptidoglycan interpeptide bridge formation. FemX catalyzes the transfer of L-Ala to the side chain of the ε-amino group of L-Lys of the peptidoglycan precursor UDP-MurNAc-pentapeptide (20). The core of the FemX protein also is composed of two GNAT domains that are related by pseudo-twofold symmetry (rmsd for internal superposition, 2.0 Å) (21). Structural domains 1 and 2 are separated by an extended cleft of 20 Å that is 15 Å deep to accommodate both the UDP-MurNAc-pentapeptide on GNAT domain 1 and the unpaired CCA acceptor arm of the charged tRNA on GNAT domain 2 (20). The superposition of GNAT domains 2 from A-PGS and FemX revealed an rmsd of 1.39 Å (GNAT domains 2 from L-PGS and FemX, 1.36 Å).

Notably, FemX was cocrystallized with a stable aminoacyl-tRNA analog. The peptidyl-RNA conjugate used mimics the peptide substrate in parallel with the instable Ala-tRNA^Ala cosubstrate, resulting in a structure that resolves the three terminal CCA nucleotides of the unpaired tRNA acceptor arm (20).

Structural superposition of the L-PGS (or A-PGS) structure with the FemX/peptidyl-RNA complex clearly localizes the terminal tRNA nucleotides in a defined cavity located on GNAT domain 2 of L-PGS (see Fig. 2E) (or A-PGS). This cavity is delineated mainly by secondary structure elements comprising a series of fully conserved amino acid residues [sequence excerpt; fully conserved residues are in bold, and residues in van der Waals distance are underscored: helix 6 and following loop region (SDAWL_{713 A-PGS}/SDEWL_{688 L-PGS}), strand I and following loop region (DLMRVHPDAPKLTM_{778 A-PGS}/DLMRYSKKAPKGIM_{742 L-PGS}), and helix 10 (LRRFK_{840 A-PGS}/FSGLRSFK_{814 L-PGS})].

Notably, Phe839 and Lys840 of A-PGS (or Phe813 and Lys814 of L-PGS) are located in a spatial position identical to that of the catalytically relevant residues Phe304 and Lys305 of FemX. Therefore, the proposed CCA-binding mode was substantiated further by mutagenesis of related A-PGS residues. Mutational analyses revealed complete A-PGS inactivation with the mutagenesis of Lys840 and Phe839, as is consistent with related results for FemX mutagenesis (20). Therefore, we concluded that Lys840 and Phe839 have a central role in tRNA substrate interaction (compare Fig. S1 and Table S2).

Table S2.

Comparison of conserved amino acid residues of A-PGS, L-PGS, and FemX indicating related enzymatic activities of mutant proteins and the inhibition of L-PGS in the presence of L-lysine amide

A-PGS residue	Related L-PGS residue (FemX residue)	Mutation	In vivo activity	Relative in vitro activity ± SD, %
Wild type			+++	100
Coordination of aminoacyl moiety of the tRNA substrate
S709	S684	S709A*	+++	5 ± 1
		S709N*	+++	17 ± 3
E720	E693	E720Q*	+++	10 ± 3
Y732	Y705	Y732A*	+	3 ± 1
D765	D739	D765A*	—	0
		D765N*	—	0
R768	R742 (Y199)†	R768S*	—	0
		R768Q*	—	0
Coordination of 3′ CCA end of tRNA
W712	W687
V769	Y733 (S259)†
A773	A737 (T264)†
K775	K739
M778	M752	M778A*	+++	50 ± 2
L836	S808 (L301)†
F839	F813 (F304; F304A: 6%; F304L: 3%)†	F839A	—	0
		F839L	+	6 ± 1
K840	K814 (K305; K305A: <0.01%; K305M: 0.5%; K305R: 0.1%)†	K840S*	—	0
		K840Q*	—	0
Coordination of the acceptor stem
K676	A649 (K169)†	K676S	++	29 ± 8
		K676E	—	0
N683	N656	N683D	++	11 ± 1
		N683S	+++	71 ± 2
R684	R657 (R176)†	R684S	++	21 ± 1
		R684E	—	0
R687	R660 (R179)†	R687S	++	13 ± 1
Conserved residues of the PG tunnel
D606	D581
G609	G584
Q636	Q611	Q636R	+	13 ± 3
		Q636W	+	10 ± 3
E640	E615
E658	E633	E658R	+	6 ± 3
		E658W	+++	23 ± 4
S763	S737	S763N*	+++	11 ± 1
G800	G774
M801	M775
R850	K824
L-PGS
Wild type				100
Wild type, 0.1 mM L-lysine amide				80
Wild type, 1 mM L-lysine amide				75
Wild type, 10 mM L-lysine amide				20

Standard in vivo and in vitro A-PGS assays have been described in ref. 15. The specific activity observed in the in vitro activity assay for wild-type A-PGS was set as 100%, and all other values of mutant proteins were related to this value. A-PG formation in the in vivo assay was classified as: +++, activity comparable with wild-type enzyme; ++, slightly decreased activity compared with wild-type enzyme; +, decreased activity compared with wild-type enzyme; —, no detectable A-PGS activity. For L-PGS inhibition studies 10 µM of purified enzyme were supplemented with L-lysine amide, and L-PG synthesis was analyzed analogously (15).

^*Data from Hebecker et al., 2011 (15).

^†Data from Fonvielle et al., 2013 (20).

In addition, residues Asp765 and Arg768 were identified as key catalytic A-PGS residues (yellow sticks in Fig. 2C). Remarkably, the L-PGS/FemX structural superposition depicted in Fig. 2E places the 3′ hydroxyl group of the terminal ribose moiety only 1.0 Å away from the cocrystallized L-lysine amide molecule (offset to the amide nitrogen of LYN, indicated by an asterisk in Fig. 2E). This compound (electron density depicted in Fig. S3) is a weak competitive inhibitor of L-PG synthesis (see Table S2). These structural and mutational analyses suggest a conserved binding mode for the unpaired CCA of the acceptor arm in aa-PGS and FemX proteins (compare Fig. 2E). The lower part of the L-lysine amide inhibitor-binding pocket is lined with a series of highly conserved amino acid residues, which are involved in an identical 3D network of polar interactions in the A-PGS and L-PGS structure (A-PGS/L-PGS: Ser709/Ser684, Glu720/Glu693, Tyr732/Tyr705, Asp765/Asp739, and Arg768/Arg742) (compare enlarged views in Fig. 2 D and G). Functional relevance for all these polar residues has been confirmed by mutagenesis as summarized in Table S2 (15). Combined interaction of Tyr732/Tyr705 and Asp765/Asp739 with the substrate α-amino group suggests an important role in the recognition of the substrate aminoacyl linkage. With regard to the catalyzed transesterification, residues Glu720/Glu693 and Ser709/Ser684 play a fundamental role in positioning Arg768/Arg742, which interacts directly with the α-carbonyl group of the cocrystallized inhibitor (compare Fig. 2 D and G). Structural and biochemical data suggest a nucleophilic attack of the 3′ hydroxyl group of PG on the Arg768/Arg742-activated α-carbonyl carbon of aminoacyl-tRNA.

Fig. S3.

Electron density map of L-lysine amide in the L-PGS structure. A feature-enhanced electron density map (45) at 1.5 σ is shown as black mesh. An Fo-Fc omit density map at 2.1 σ is shown in magenta.

Specific synthesis of A-PG or L-PG requires accurate recognition of Ala-tRNA^Ala versus Lys-tRNA^Lys. Inspection of the upper part of the aminoacyl-binding cavity of the L-PGS and A-PGS structure did not reveal any conserved or supplemental amino acid residues as direct determinants for the specific recognition (or steric exclusion) of the lysyl- versus the much smaller alanyl-substrate moiety. The four-aminobutyl side chain of the cocrystallized L-lysine amide is solely in contact with main-chain L-PGS atoms, which were found in an almost identical position in the related A-PGS structure (compare Fig. 2 D and G). One might argue that the sole amino acid moiety of the tRNA substrate does not account for the overall specificity of aa-PGS enzymes. Accordingly, we investigated whether the tRNA portion was a main aa-PGS determinant by using tRNA^Lys that was mischarged with the smaller amino acid alanine.

The highly specific A-PGS or L-PGS enzyme was analyzed in the presence of tRNA^Lys that was mischarged with the amino acid alanine (Ala-tRNA^LysC70U; Fig. 3). The specific requirement for the C70U mutation within the synthesis of Ala-tRNA^LysC70U is illustrated in the sequence comparison of the tRNA species we used, depicted in Fig. S4. This artificial Ala-tRNA^LysC70U substrate was not accepted by L-PGS (compare Fig. 3 and Fig. S5). In contrast, the A-PGS enzyme revealed a relative activity of 80% compared with the natural Ala-tRNA^Ala substrate (Fig. 3). These experiments might indicate that the tRNA moiety in combination with the amino acid moiety of Ala-tRNA^Ala is relevant for A-PGS substrate recognition. These findings are supported further by earlier experiments (using artificial tRNA microhelices), which revealed the five terminal base pairings as important elements of A-PGS substrate recognition (15).

Fig. 3.

A-PG synthesis of A-PGS and L-PGS using mischarged ¹⁴C-Ala-tRNA^LysC70U. (A) ¹⁴C-Ala-tRNA^Ala or ¹⁴C-Ala-tRNA^LysC70U was used as substrate of A-PGS (0.2 µM) and L-PGS (1 µM) in combination with PG. Control reactions with purified L-PGS in the presence of the Lys-tRNA^Lys substrate revealed efficient L-PG formation (compare with Fig. S5). Synthesis of aa-PG was analyzed by lipid extraction and liquid scintillation analysis. (B) Sequence of Ala-tRNA^Ala from B. licheniformis. (C) Sequence of Ala-tRNA^LysC70U from P. aeruginosa. The mutated base is indicated by an asterisk.

Fig. S4.

Comparison of the tRNA acceptor stem of Ala-tRNA^Ala, Lys-tRNA^Lys, and misacylated Ala-tRNA^LysC70U, identity elements of aminoacyl-tRNA synthetases and substrate recognition of A-PGS and L-PGS. The acceptor stems of Ala-tRNA^Ala from B. licheniformis and Lys-tRNA^Lys from P. aeruginosa and from misacylated Ala-tRNA^LysC70U (point mutation indicted by asterisk) are shown. Enzymatic misacylation of tRNA^Ala or tRNA^Lys requires the respective tRNA identity elements (highlighted in gray) of alanyl-tRNA synthetase [AlaRS; A73 discriminator base, G3-U70 (37, 46)] or lysyl-tRNA synthetase [LysRS; A73 discriminator base, G3-C70, UUU34-36 anticodon (36, 47)], respectively. Sequence differences for tRNA^Ala and tRNA^Lys are highlighted by blue boxes. Substrate recognition of aa-PGS was analyzed using the artificial substrate Ala-tRNA^LysC70U. Synthesis of this misacylated tRNA by AlaRS requires a single base mutation (36). Ala-tRNA^LysC70U was not accepted by L-PGS but was efficiently converted by A-PGS. Notice that tRNA^LysC70U is not accepted as a substrate of LysRS because of the C70U mutation (36).

Fig. S5.

L-PG synthesis of L-PGS and A-PGS using ¹⁴C-Lys-tRNA^Lys as substrate. Isolated ¹⁴C-Lys-tRNA^Lys was used as substrate in the presence of purified A-PGS (1 µM) or L-PGS (1 µM) in the presence of PG. L-PG synthesis was analyzed by lipid extraction and liquid scintillation counting.

The theoretical comparison of the tRNA^Ala and the tRNA^Lys acceptor stem sequences shown in Fig. S4 reveals only base pairings G4–C69/U4–A69, C6–G67/G6–C67, and U70/C70 for the discrimination of tRNA^Ala/tRNA^Lys. However, the observed A-PGS activity in the presence of tRNA^LysC70U indicates that G4, C69, C6, and G67 are nondiscriminating bases of A-PGS substrate recognition. Therefore, we concluded that U70 has a relevant role in A-PGS substrate recognition. In addition, specific substrate discrimination also might include dynamic interaction among aa-PGS and the overall tRNA substrate as exemplified for other transient protein/tRNA complexes (22, 23). Future cocrystallization experiments in the presence of sophisticated nonhydrolyzable aminoacyl-tRNA analogs might reveal such conformational rearrangements with relevance for the specific recognition of Ala-tRNA^Ala or Lys-tRNA^Lys.

Inspection of the protein surface of A-PGS and L-PGS revealed a series of conserved amino acid residues located on helix 5 (Lys676, Arg684, Arg687, and Asn683) as candidates for additional tRNA interaction. Mutagenesis of these residues into small polar residues always resulted in moderately retained activities (K676S, 29%; R684S, 21%; R687S, 13%; N683S, 71%; and N683D, 11%). In contrast, the individual charge reversal caused by glutamate insertion resulted in a complete loss of A-PGS activity, indicating a potential tRNA interaction. Subsequently, a representative tRNA structure from the Protein Data Bank (PDB ID code 1TN1) was positioned manually onto the structure of A-PGS and L-PGS using the aminoacyl-binding pocket and the CCA-binding pocket as points of reference. This experiment revealed the mutagenized amino acid residues of helix 5 (Lys676, Arg684, Arg687) in van der Waals distance from the previously determined recognition elements of the tRNA^Ala acceptor stem (base pairings G2–C71, G3–U70, and G4–C69). Fig. 2F shows an enlarged view of this A-PGS/tRNA acceptor stem interaction, and the overall models of tRNA/A-PGS and tRNA/L-PGS interaction are depicted in Fig. S6.

Fig. S6.

Model for L-PGS and A-PGS tRNA interaction. A tRNA molecule (PDB ID code 1TN1) was manually positioned onto the structure of L-PGS (A) and A-PGS (B) using the aminoacyl-binding pocket and the CCA-binding pocket as reference points. Substitution of amino acid residues Lys676, Arg684, and Arg687 located on helix 5 results in impaired A-PGS activity. The terminal base pairings G2–C71, G3–U70, and G4–C69 have been identified previously as tRNA recognition elements (compare Table S2 and ref. 15). The theoretical position of the C70 or U70 base in the model of L-PGS or A-PGS is indicated by an asterisk.

Aa-PGS catalysis requires the recruitment of the hydrophobic lipid substrate, whereas FemX binds the polar UDP-MurNAc-pentapeptide in an extended cleft located on GNAT domain 1 (20). Accordingly, we propose a completely unrelated aa-PGS substrate-binding mode with respect to PG. A comparison of the overall structure of aa-PGS and FemX reveals that the core secondary structural elements of GNAT domain 1 are tilted by up to 6 Å toward the domain interface, thereby closing the UDP-MurNAc-pentapeptide–binding cavity of FemX by virtue of helix 1 (PDGGLALT₅₇₇) and loop region ARRGRSMI₆₀₂ and also by an insertion (loop region EKGFSLGR₇₂₇) and the extension (LIAGGLTGL₈₇₈) located at the C terminus of A-PGS.

In Fig. 2 C and D an intersection of both aa-PGS proteins reveals a 23-Å tunnel with a diameter of 4–8 Å, which is connected directly to the aminoacyl-tRNA– binding cavity of A-PGS and L-PGS, respectively. This theoretical “backdoor entrance” is delineated by conserved residues of GNAT domains 1 and 2 (see Fig. S1) and provides an elegant solution for the aa-PGS catalysis at the water–lipid interface: The water-insoluble lipid substrate was placed into the lower part of the A-PGS tunnel by running a molecular docking simulation (Autodock Vina software in combination with PG C5:0/C8:0; PDB ligand ID code AGA). The relative positioning of the PG molecule used in the postulated lipid-binding site of A-PGS is depicted in Fig. 2H. The 3′ hydroxyl of the polar lipid head group protrudes toward the aminoacyl-binding site, whereas the branched diacylglycerol moiety appears to have a high degree of flexibility concerning the recognition of the respective fatty acid moieties of PG [C16:0/C19:0 cis 9, 10 cyclopropane predominant in P. aeruginosa (5, 24)]. This theoretical docking mode is in agreement with earlier experiments investigating A-PGS activity in the presence of artificial PG substrates (15). Modification of the respective fatty acid alkyl chains by (i) variation of saturation, (ii) fourfold methylation, (iii) using a monoacylated PG derivative, or (iv) using a short-chain C6 fatty acid PG variant did not hamper the activity of A-PGS. However, the alteration of the polar head group into an ethylene glycol moiety or by using diphosphatidylglycerol with a symmetrically branched glycerol moiety did not result in any detectable A-PGS activity, indicating that the polar head group of PG is the dominant determinant of lipid substrate recognition (15).

The proposed PG substrate-binding mode was analyzed experimentally by site-directed mutagenesis of solvent-exposed amino acid residues located at the bottleneck that forms the connection to the aminoacyl-tRNA–binding site of A-PGS (see Fig. 2C and the close-up view depicted in Fig. 2H). Replacing residues Gln636, Glu658, and Ser763 (highlighted in light gray in Fig. 2H) with significantly bulkier side chains (tryptophane, arginine, or asparagine) resulted in only moderately retained activities of 6% for E658R, 13% for Q636R, 10% for Q636W, 23% for E658W, and 11% for S763N (Table S2). These biochemical and structural findings suggest the binding of the polar aminoacyl-tRNA molecule opposite the hydrophobic lipid substrate as a fundamental principle for the aa-PGS catalysis at the water–lipid interface. The elucidated modes of substrate recognition provide a framework for the future development of aa-PGS inhibitors as a new strategy to render pathogenic bacteria more susceptible to established antibiotics and also to the wide range of naturally occurring antimicrobial defense molecules of the human host.

Materials and Methods

Production and Purification of P. aeruginosa A-PGS and B. licheniformis L-PGS.

Base pairs 1627–2643 of ORF PA0920 from P. aeruginosa PAO1 and base pairs 1555–2550 of ORF yfiX from B. licheniformis DSM13 were PCR-amplified using oligonucleotide pairs 1 and 2 and 3 and 4 (Table S3), respectively, and were cloned into the XmaI and SacI sites of pET52b(+) (Novagen) for expression with a cleavable N-terminal Strep-II-tag. The SerP (surface entropy reduction) server (25) was used to identify mutations that may facilitate optimized crystallization of A-PGS (KGKE₆₇₄ to AGAA₆₇₄). To exchange amino acid residues of the catalytic domain of A-PGS, the QuikChange kit (Agilent) was used according to the manufacturer’s instructions in combination with oligonucleotide pairs 5 and 6 to 31 and 32.

Table S3.

Oligonucleotides used in this study

Primer	5′ – 3′ sequence
1	TTAACCCGGGCGCGCGGCACC
2	GAGTGAGCTCTCAGCGTTTCACCAATC
3	GCGCCCCGGGAACAGGAAAACGAAAGAGATC
4	GCGCGAGCTCTCATTTCGTTCTCCTTCC
5	CCGGTGTTCTACTGGGTGCGTGCCGAGAAC
6	GTTCTCGGCACGCACCCAGTAGAACACCGG
7	CCGGTGTTCTACCGTGTGCGTGCCGAGAAC
8	GTTCTCGGCACGCACACGGTAGAACACCGG
9	CAAGCTCGGCGAACGTGCGCGAGTC
10	GACTCGCGCACGTTCGCCGAGCTTG
11	CCTCAAGCTCGGCGAATGGGCGCGAGTCGAC
12	GTCGACTCGCGCCCATTCGCCGAGCTTGAGG
13	GGGCAAGGAAATGGAAGACCTGCGCTACAC
14	GTGTAGCGCAGGTCTTCCATTTCCTTGCCC
15	GAACAAGGGCAAGGAAATGTCTGACCTGCGCTACACCTG
16	CAGGTGTAGCGCAGGTCAGACATTTCCTTGCCCTTGTTC
17	GCTACACCTGGGACCGCGGCCAGC
18	GCTGGCCGCGGTCCCAGGTGTAGC
19	CTGCGCTACACCTGGTCTCGCGGCCAGCGCG
20	CGCGCTGGCCGCGAGACCAGGTGTAGCGCAG
21	GCGCTACACCTGGAACGAAGGCCAGCGCGACG
22	CGTCGCGCTGGCCTTCGTTCCAGGTGTAGCGC
23	GCGCTACACCTGGAACTCTGGCCAGCGCGACG
24	CGTCGCGCTGGCCAGAGTTCCAGGTGTAGCGC
25	GGAACCGCGGCCAGTCTGACGGCCTGGCC
26	GGCCAGGCCGTCAGACTGGCCGCGGTTCC
27	CAGGGGCTGCGACGCGCTAAGGACAAGTTCCAG
28	CTGGAACTTGTCCTTAGCGCGTCGCAGCCCCTG
29	CAGGGGCTGCGACGCCTGAAGGACAAGTTCCAG
30	CTGGAACTTGTCCTTCAGGCGTCGCAGCCCCTG
31	GCTGCGCTTCGACCTGGAGAACGCTGGCGCTGCTATGAAAGACCTGCGCTACACCTG
32	CCAGGTGTAGCGCAGGTCTTTCATAGCAGCGCCAGCGTTCTCCAGGTCGAAGCGCAC
33	AATTCTAATACGACTCACTATAGGGGCCTTAGCTCAGCTGGGAGAGCGCCTGCTTTGCACGCAGGAGGTCAGCGGTTCGATCCCGCTAGGCTCCACCAG
34	GATCCTGGTGGAGCCTAGCGGGATCGAACCGCTGACCTCCTGCGTGCAAAGCAGGCGCTCTCCCAGCTGAGCTAAGGCCCCTATAGTGAGTCGTATTAG
35	AATTCTAATACGACTCACTATAGGGTCGTTAGCTCAGTCGGTAGAGCAGTTGGCTTTTAACCAATTGGTCGTAGGTTCGAATCCTACACGATCCACCAG
36	GATCCTGGTGGATCGTGTAGGATTCGAACCTACGACCAATTGGTTAAAAGCCAACTGCTCTACCGACTGAGCTAACGACCCTATAGTGAGTCGTATTAG
37	AATTCTAATACGACTCACTATAGGGTCGTTAGCTCAGTCGGTAGAGCAGTTGGCTTTTAACCAATTGGTCGTAGGTTCGAATCCTACACGACCCACCAG
38	GATCCTGGTGGGTCGTGTAGGATTCGAACCTACGACCAATTGGTTAAAAGCCAACTGCTCTACCGACTGAGCTAACGACCCTATAGTGAGTCGTATTAG

Restriction sites are shown in italics. Bold fonts indicate exchanged nucleotides for mutagenesis of the catalytic domain of A-PGS from P. aeruginosa.

A-PGS and L-PGS genes were expressed, and recombinant proteins were purified to apparent homogeneity as follows: Transformed Escherichia coli BL 21(DE3) cells for the production of A-PGS_543–881 and Tuner (DE3) cells for production of L-PGS_519–850 were cultivated at 37 °C in LB medium supplemented with 100 µg/mL ampicillin. At an OD₅₇₈ of 0.5, protein production was induced with 50 µM of isopropyl β-d-1-thiogalactopyranoside, and cells were shifted to 17 °C for 18 h. Selenomethionine-labeled surface mutant A-PGS_543–881 AGAA with 90% SeMet occupancy was produced as described elsewhere (26).

Cells were harvested by centrifugation and disrupted by a French press at 19,200 psi in lysis buffer [100 mM Tris⋅HCl (pH 7.5), 400 mM NaCl, 20 mM MgCl₂, 5% (wt/vol) glycerol, 2 mM DTT].

After ultracentrifugation for 1 h at 110,000 × g at 4 °C, the supernatant was applied to 1 mL of Strep-Tactin Superflow resin (IBA), which was equilibrated with lysis buffer. Following washing with 10 mL of lysis buffer and 10 mL of elution buffer 1 [20 mM Tris⋅HCl (pH 7.5), 200 mM NaCl, 0.5 mM MgCl₂, 5% (wt/vol) glycerol, 2 mM DTT] for A-PGS_543–881 AGAA and A-PGS_543–881, or alternatively elution buffer 2 [20 mM Tris⋅HCl (pH 7.5), 300 mM NaCl, 5 mM MgCl₂, 5% (wt/vol) glycerol, 2 mM DTT] for L-PGS_519–850, the proteins were liberated and eluted from the resin by cleavage of the Strep-II-tag via PreScission protease treatment (GE Healthcare). The GST-tagged protease was removed using Glutathione Sepharose 4FF (GE Healthcare). The elution fractions containing A-PGS and L-PGS proteins, respectively, were concentrated to ∼10 mg/mL using Vivaspin 15 centrifugal concentrators with a 10-kDa cutoff (Sartorius).

Protein Crystallization.

Crystals were obtained in hanging-drop vapor-diffusion experiments at 4 °C by mixing 2 µL of protein with 2 µL of reservoir solution. Crystals of selenomethionine-labeled A-PGS_543–881 AGAA grew from a solution containing 7.5 mM CoCl₂, 85 mM Mes (pH 5.7), 1.53 M (NH₄)₂SO₄, and 15% (vol/vol) glycerol. Crystals of native A-PGS_543–881 AGAA grew from 85 mM Na-acetate (pH 6.37), 1.6 M (NH₄)₂SO₄, and 20% (vol/vol) glycerol. Crystals of L-PGS_519–850 were obtained from a solution of 0.2 M NaCl, 0.1 M phosphate-citrate (pH 4.2), and 10% (wt/vol) PEG3000 supplemented with 0.5 mM L-lysine amide. Needle-shaped crystals grew within 1–2 wk. Crystals were shock-cooled in liquid nitrogen. L-PGS_519–850 crystals were cryoprotected with 30% (vol/vol) glycerol in reservoir solution before cooling in liquid nitrogen.

Data Collection, Structure Determination, and Refinement.

Diffraction data of A-PGS and L-PGS crystals were collected on beamline 14.2 (27) of the Berlin Electron Storage Ring Society for Synchrotron Radiation (BESSY) II electron storage ring (Berlin-Adlershof) and on beamline P11 (28) of the Positron-Electron Tandem Ring Accelerator (Petra) III at DESY (Hamburg, Germany). Integration and space group assignment were carried out with XDS (29). A crystal of selenomethionine-derivatized A-PGS was used for data collection at the absorption edge of selenium. The resulting anomalous signal was sufficient to obtain experimental phases in a single-wavelength anomalous dispersion experiment and to compute initial electron density with phenix.autosol (30). A first structural model was built by phenix.autobuild (30), which subsequently was improved by manual rebuilding in COOT (31) and refinement in phenix.refine (30). An improved A-PGS model was obtained through refinement against high-resolution data of a native A-PGS crystal. Diffraction data of L-PGS crystals showed significant anisotropy and were subjected to anisotropy correction using the anisotropy correction server (32). The phase problem was solved by molecular replacement with phaser (33) using a pruned A-PGS monomer as search model. Phaser placed two monomers in the asymmetric unit with reasonable confidence. However, the resulting electron density was poor. Hence, model and density were improved by alternate rebuilding and relaxation cycles with phenix.mr_rosetta (34) and then finalized through manual rebuilding and refinement. The complete data collection and refinement statistics are shown in Table S1.

A-PGS/L-PGS Enzyme Assay.

The in vivo A-PGS and L-PGS activities were determined as described before (15). To validate the purified (mutant) proteins, our well-established in vitro activity assay was performed (15). In brief, an E. coli strain overproducing either alanyl-tRNA synthetase (15) or lysyl-tRNA synthetase (35) provided the substrate molecules PG and aminoacylated tRNA^Ala or tRNA^Lys. The formation of A-PG or L-PG, respectively, was determined by liquid scintillation counting using radioactively labeled [1-¹⁴C]-L-alanine (51 mCi/mmol; Moravek Biochemicals) or [U-¹⁴C]-L-lysine (288 mCi/mmol; Moravek Biochemicals) (15).

Preparation, Purification, and Aminoacylation of RNA Transcripts.

The tRNA^Ala gene from B. licheniformis, the tRNA^Lys gene from P. aeruginosa, and the sequence for tRNA^LysC70U (carrying identity elements for misacylation with alanine by alanyl-tRNA synthetase; G3-C70 base pair mutated to G3-U70) were cloned into the pUC18 vector (using oligonucleotides 33–38). In vitro-transcribed tRNAs were purified via MonoQ chromatography, folded, and acylated with [1-¹⁴C]-L-alanine (for tRNA^Ala) or [U-¹⁴C]-L-lysine (for tRNA^Lys), as described elsewhere (15, 36).

Misacylation of tRNA^Lys Using Alanyl-tRNA Synthetase.

Substrate recognition of alanyl-tRNA synthetase is one of the rare instances in which only a single base pair of the acceptor stem acts as a major identity element of tRNA recognition (37). Therefore, native tRNA^Lys from P. aeruginosa functions as a substrate of alanyl-tRNA synthetase because of the sole mutation of the base pairing G3-C70 into G3-U70 (36). Hence, in vitro synthesis of the misacylated Ala-tRNA^LysC70U offers an alternative methodology to investigate aa-PGS substrate recognition. In vitro-transcribed tRNA^LysC70U was misacylated efficiently in the presence of 0.42 µM of ¹⁴C-Ala and 1 µM E. coli alanyl-tRNA synthetase (15). Synthesis of a related Lys-tRNA^LysC70U is strongly hampered because a lysyl-tRNA synthetase revealed a loss of activity by a factor of >1,000 as the result of a G3-U70 mutation (36).

A-PGS/L-PGS Activity Assays in the Presence of Ala-tRNA^Ala and Misacylated Ala-tRNA^LysC70U.

We used 0.42 µM of ¹⁴C-Ala-tRNA^Ala or ¹⁴C-Ala-tRNA^LysC70U as substrate for A-PGS (0.2 µM) and L-PGS (0.1, 1 µM, and 10 µM) in the presence of 2 mg/mL PG (Sigma-Aldrich) supplemented with 1.76 mg/mL Triton X-100 in the respective elution buffers. Synthesis of A-PG was analyzed by lipid extraction and liquid scintillation analysis. Control experiments using ¹⁴C-Lys-tRNA^Lys were performed to demonstrate L-PGS activity. This experimental setup was used to elucidate the overall contribution of the tRNA substrate moiety.

Structure-Based Sequence Analysis.

Structure-based sequence analyses were calculated by the MatchMaker and MatchAlign subroutines of UCSF Chimera (38, 39). Figures were prepared with Pymol (40) and UCSF Chimera. Rmsds were calculated in Pymol.

Molecular Docking of PG.

Docking calculations were performed by means of Autodock Vina (41) as part of the MGLTools package (mgltools.scripps.edu/) to investigate the binding mode of the lipid substrate. The PG ligand (PG C5:0/C8:0, PDB ligand ID AGA, from structure 1q16; ref. 42) was extracted from PDB, charges and rotational bonds were assigned, and no flexibility was allowed for side-chain residues for the A-PGS structure. The search volume was assigned first to the whole A-PGS and then stepwise, limited to the lower part of the substrate tunnel to exclude docking to the outer shell of the molecule. All conformers showing diametrically opposed fatty acid moieties were considered biologically irrelevant. The theoretical lipid-binding mode depicted in Fig. 2H shows the best result, having the lowest binding energy of the remaining list.