| Abstract: | ![]()
To identify lipids with roles in tuberculosis disease, we systematically compared the lipid content of virulent Mycobacterium tuberculosis with the attenuated vaccine strain Mycobacterium bovis bacillus Calmette–Guérin. Comparative lipidomics analysis identified more than 1,000 molecular differences, including a previously unknown, Mycobacterium tuberculosis-specific lipid that is composed of a diterpene unit linked to adenosine. We established the complete structure of the natural product as 1-tuberculosinyladenosine (1-TbAd) using mass spectrometry and NMR spectroscopy. A screen for 1-TbAd mutants, complementation studies, and gene transfer identified Rv3378c as necessary for 1-TbAd biosynthesis. Whereas Rv3378c was previously thought to function as a phosphatase, these studies establish its role as a tuberculosinyl transferase and suggest a revised biosynthetic pathway for the sequential action of Rv3377c-Rv3378c. In agreement with this model, recombinant Rv3378c protein produced 1-TbAd, and its crystal structure revealed a cis-prenyl transferase fold with hydrophobic residues for isoprenoid binding and a second binding pocket suitable for the nucleoside substrate. The dual-substrate pocket distinguishes Rv3378c from classical cis-prenyl transferases, providing a unique model for the prenylation of diverse metabolites. Terpene nucleosides are rare in nature, and 1-TbAd is known only in Mycobacterium tuberculosis. Thus, this intersection of nucleoside and terpene pathways likely arose late in the evolution of the Mycobacterium tuberculosis complex; 1-TbAd serves as an abundant chemical marker of Mycobacterium tuberculosis, and the extracellular export of this amphipathic molecule likely accounts for the known virulence-promoting effects of the Rv3378c enzyme.With a mortality rate exceeding 1.5 million deaths annually, Mycobacterium tuberculosis remains one of the world''s most important pathogens (1). M. tuberculosis succeeds as a pathogen because of productive infection of the endosomal network of phagocytes. Its residence within the phagosome protects it from immune responses during its decades long infection cycle. However, intracellular survival depends on active inhibition of pH-dependent killing mechanisms, which occurs for M. tuberculosis but not species with low disease-causing potential (2). Intracellular survival is also enhanced by an unusually hydrophobic and multilayered protective cell envelope. Despite study of this pathogen for more than a century, the spectrum of natural lipids within M. tuberculosis membranes is not yet fully defined. For example, the products of many genes annotated as lipid synthases remain unknown (3), and mass spectrometry detects hundreds of ions that do not correspond to known lipids in the MycoMass and LipidDB databases (4, 5).To broadly compare the lipid profiles of virulent and avirulent mycobacteria, we took advantage of a recently validated metabolomics platform (4). This high performance liquid chromatography–mass spectrometry (HPLC-MS) system uses methods of extraction, chromatography, and databases that are specialized for mycobacteria. After extraction of total bacterial lipids into organic solvents, HPLC-MS enables massively parallel detection of thousands of ions corresponding to diverse lipids that range from apolar polyketides to polar phosphoglycolipids. Software-based (XCMS) ion finding algorithms report reproducibly detected ions as molecular features. Each feature is a 3D data point with linked mass, retention time, and intensity values from one detected molecule or isotope. All features with equivalent mass and retention time from two bacterial lipid extracts are aligned, allowing pairwise comparisons of MS signal intensity to enumerate molecules that are overproduced in one strain with a false-positive rate below 1% (4).This comparative lipidomics system allowed an unbiased, organism-wide analysis of lipids from M. tuberculosis and the attenuated vaccine strain, Mycobacterium bovis Bacillus Calmette–Guérin (BCG). BCG was chosen because of its worldwide use as a vaccine and its genetic similarity to M. tuberculosis (6). We reasoned that any features that are specifically detected in M. tuberculosis might be clinically useful as markers to distinguish tuberculosis-causing bacteria from vaccines. Furthermore, given the differing potential for productive infection by the two strains, any M. tuberculosis-specific compounds would be candidate virulence factors. Comparative genomics of M. tuberculosis and BCG successfully identified “regions of deletion” (RD) that encode genes that were subsequently proven to promote productive M. tuberculosis infection (7), including the 6-kDa early secreted antigenic target (ESAT-6) secretion system-1 (ESX-1) (8, 9). We reasoned that a metabolite-based screen might identify new virulence factors because not all functions of RD genes are known. Also, biologically important metabolites could emerge from complex biosynthetic pathways that cannot be predicted from single-gene analysis.Comparison of M. tuberculosis and BCG lipid profiles revealed more than 1,000 differences, among which we identified a previously unknown M. tuberculosis-specific diterpene-linked adenosine and showed that it is produced by the enzyme Rv3378c. Previously, Rv3378c was thought to generate free tuberculosinol and isotuberculosinol (10–12). This discovery revises the enzymatic function of Rv3378c, which acts as a virulence factor to inhibit phagolysosome fusion (13). Whereas current models of prenyl transferase function emphasize iterative lengthening of prenyl pyrophosphates using one binding pocket, the crystal structure of Rv3378c identifies two pockets in the catalytic site, establishing a mechanism for heterologous prenyl transfer to nonprenyl metabolites. |
