Mapping unsolved lipidomes accelerates lipid discovery in major bacterial pathogens

Abstract

Unlike gene-first approaches to understanding bacterial pathogenesis, molecule-forward discovery can uncover unexpected chemical diversity. Here, new lipidomic analytical methods and quality metrics defined the large scope of unknown lipids in the world's deadliest pathogen, Mycobacterium tuberculosis (Mtb). This map allowed rapid discovery of Mtb lysyldiacylglycerol linked to the biosynthetic gene lysX, which controls in vivo infection outcomes in moth larvae, mice, guinea pigs, and here, zebrafish. A broader search for orthologous lysyltransferase domains identified the Staphylococcus aureus virulence gene mprF, where the same lipoamino acid was shown to be a previously unknown biosynthetic product. Thus, lipidomic mapping showed that the cell envelope composition of well-studied bacterial pathogens remains substantially unsolved and offers a new way to generate lists of discoverable lipids to accelerate molecular discovery.

Fig. 1.. Mapping a lipidome of unknowns in M. tuberculosis

(A) The Mtb cell envelope comprises two lipid membranes. (B) The 5542 features of the positive mode lipidome of Mtb H37Rv in biological quadruplicate underwent sequential, semi-automated censoring to remove (C) solvent ions, (D) salt clusters (E) isotopes and alternate adducts, and (F) in source multimers and fragments, yielding 588 credentialed features. (G) Source databases were updated and combined to create a metabolite and lipid database of defined m/z values (MycoMassDB), and a propagated database using LOBSTAHs (21) to generate theoretical lipid variants. (H) The 588 qualified features of the ‘detectable lipidome’ were classified as (I) 276 features of the ‘lipidome of knowns’, with lead compounds identified by collisional MS and (J) 312 features of the ‘lipidome of unknowns’ grouped into lipid families and ranked as high, medium, or low confidence ions based on four criteria: detectable acylforms, the presence of alternate adducts, detection in reversed phase chromatography, and alternate solvent extraction. (K) The lipidomes of detectable, credentialed known and unknown lipids were summarized. PDIM, phthiocerol dimycocerosates; Ac4SGL, tetra-acyl sulfoglycolipid; Ac3SGL, triacyl sulfoglycolipid; Ac2SGL, diacyl sulfoglycolipid; MK, menaquinone; PAT, polyacyl trehalose; TAT, triacyl trehalose; DAT, diacyl trehalose; DAG, diacylglycerol; TAG, triacylglycerol; CL, cardiolipin; PI, phosphatidylinositol; PE, phosphatidylethanolamine; LPE, lyso phosphatidylethanolamine; TMM, trehalose monomycolate; GroMM, glycerol monomycolate; Ac2PIM2, diacyl phosphatidylinositol dimannoside; Ac1PIM2 monoacyl phosphatidylinositol dimannoside; AcXPIM1, acyl phosphatidylinositol monomannoside; PIM2, phosphatidylinositol dimannoside. *Lipids with asterisks mass matched to MycoMassDB and were not studied by collisional MS.

Fig. 2.. Identification of lysX-dependent lysyldiacylglycerol.

(A) A cluster of high confidence (blue) features was shown to have 7 acylforms. (B) Collisional MS identified lysyldiacylglycerol (lysylDAG), with (C) representative mass chromatograms of acylforms in Mtb H37Rv. (D) A search for Mtb genes with homology to lysyltransferases by Interpro (49) subdomain structure identified two candidates, lysX and Rv1619. (E) Representative single ion chromatograms for the mass of lysylDAG (C34:1, [M+H]+) in total Mtb lipids from CRISPRi knockdowns of lysX and Rv1619 show lysX-dependence. (F) Chromatogram areas for summed lysylDAG acylforms, tuberculosinyl adenosine (m/z 540.355 [M+H]+, TbAd), and monoacyl phosphatidylinositol dimannoside (m/z 1432.942 [M+NH4]+, Ac1PIM2) analyzed in biological triplicate in the lysX KD background. P-value determined by pairwise t-test. (G) Comparative lipidome of lysX knockdown against the untreated control in biological triplicate, representative of two independent experiments. A significance threshold using the Benjamini-Hochberg adjusted P-value < 0.05 and two-fold change was used to identify lysX-dependent features. (H) 63 monoisotopic features were classified into three known and one unknown lipid families by grouping acylforms, m/z matching, and collisional MS. (I) Representative mass chromatograms of lysX-dependent families show the relative difference in intensity with lysX KD. (J) Both lysylDAG with lysine in the sn2 (blue) and sn3 (red) positions were synthesized in 5 steps. (K) Mass chromatograms of the C32:0 (m/z 697.609 [M+H]+) acylform of Mtb H37Rv lysylDAG (black) co-elutes in the reversed phase with the sn3 synthetic lysylDAG (red) and not the sn2 synthetic isomer (blue). (L) The z-score of the intensities of 7 alkylforms of lysylDAG, 17 diacylglycerol (DAG), and 16 phosphatidylethanolamine (PE) in Mtb H37Rv were measured across growth timepoints and plotted as scaled circles with a solvent only negative control. Distributions with a non-significant P-value of the F-statistic across all non-solvent pairwise timepoint contrasts, P > 0.01, are shown with 25% opacity. Representative of two independent experiments. Representative mass chromatograms show the distribution of (M) DAG, (N) lysylDAG and (O) PE across logarithmic to stationary growth timepoints sampled.

Fig. 3.. Lysyldiacylglycerol-lysX: a conserved mycobacterial virulence pair.

(A) Subset of a phylogenetic tree of LysX amino acid sequence shows the conservation of lysX across mycobacteria, with MprF-domain enzyme classification described previously (38). (B) Mass chromatograms show the distribution of representative lysyldiacylglycerol (lysylDAG) acylforms across mycobacteria profiled. (C) The z-score intensities of acylforms of lysylDAG, diacylglycerol (DAG), and phosphatidylethanolamine (PE) were plotted as scaled circles in biological quadruplicate lipid extracts from 3 Mtb strains and 12 mycobacteria grown in parallel. (D) Mass chromatograms of lysylDAG and DAG in M. marinum (Mm) WT, lysX transposon mutant (TnlysX) and lysX complement (TnlysX::hsp60::lysX) show lysX dependence. (E) A threshold of two-fold change and P-value < 0.01 in a compound lipidomic contrast of Mm WT and lysX complement against TnlysX identified lipids significantly changed by lysX disruption. Strains were grown in biological triplicate, representative of two independent experiments. (F) 142 monoisotopic features were enriched in Mm TnlysX. Grouping and lead compound collisional MS identified 74 acylforms of triacylglycerol (TAG) and 2 acylforms of phosphatidylethanolamine (PE). (G) 72 monoisotopic features were enriched in Mm WT and ::lysX. Grouping and lead compound collisional-MS identified 25 acylforms of DAG, 1 of TAG, and 7 of lysylDAG. Representative single ion chromatograms of (H) PE and TAG show enrichment in both TnlysX (black) and ::lysX (blue), whereas (I) DAG, lysylDAG, and TAG show enrichment in Mm WT (red) and ::lysX (blue). lysylDAG was absent in the lysX mutant. (J) Chromatographic areas in the Mm lysX lipidome were summed and significant differences were evaluated by two-way ANOVA with Tukey’s post-test (*: P < 0.05; **: P < 0.01; ***: P < 0.001; ****: P < 0.0001). (K) M. marinum burden in zebrafish infection measured using bacterial mCerulean fluorescence. Data shows one representative experiment of 3 biological replicates with 30–60 independent infections per replicate. Median and 95% confidence interval are displayed. Statistical analyses were performed using one-way Welch’s ANOVA followed by a Dunnett’s T3 multiple comparison test of each group to the WT strain (ns: P > 0.05; **: P ≤ 0.01; ***: P ≤ 0.001; ****: P < 0.0001). (L) Representative images from zebrafish, depicted in (K) in pink, infected with an initial dose of 150–200 fluorescence units of either WT or TnlysX at 4 days post infection.

Fig. 4.. Lysyldiacylglycerol in firmicutes.

(A) Amino acid sequence phylogeny of major MprF-domain containing proteins in actinobacteria and firmicutes show three clades. (B) A comparative lipidome of S. aureus 113 WT against the ΔmprF mutant was used to identify mprF-dependent features in biological triplicate (C) Grouping by transformed mass defect, 5 families with > 2-fold change were enriched in the S. aureus WT strain, Benjamini-Hochberg adjusted P value < 0.01. (D) One family consistent with a known product was identified as lysyl phosphatidylglycerol (lysylPG). (E) Representative chromatograms show the lysylPG acylform distribution S. aureus WT, left, and mass chromatogram showing absence in the mprF mutant, right. (F and H) Two previously unknown families were identified by collisional MS as an (F) hexosyl-modified lysylPG (hexlysylPG) and (H) lysyldiacylglycerol (lysylDAG). (G and I) Representative mass chromatograms show the acylform distribution of hexlysylPG and lysylDAG in S. aureus WT, left, and mass chromatogram showing absence in the mprF mutant, right, consistent with MprF products, and (J) show detection of lysylPG, lysylDAG, and DAG across growth phases. (K) A summary figure maps the identification of lysylDAG in the ‘lipidome of unknowns’ of Mtb H37Rv, extending this discovery to representative mycobacteria dependent on lysX and to S. aureus dependent on mprF.