Mycobacterium tuberculosis functional network analysis by global subcellular protein profiling

Abstract

Trends in increased tuberculosis infection and a fatality rate of approximately 23% have necessitated the search for alternative biomarkers using newly developed postgenomic approaches. Here we provide a systematic analysis of Mycobacterium tuberculosis (Mtb) by directly profiling its gene products. This analysis combines high-throughput proteomics and computational approaches to elucidate the globally expressed complements of the three subcellular compartments (the cell wall, membrane, and cytosol) of Mtb. We report the identifications of 1044 proteins and their corresponding localizations in these compartments. Genome-based computational and metabolic pathways analyses were performed and integrated with proteomics data to reconstruct response networks. From the reconstructed response networks for fatty acid degradation and lipid biosynthesis pathways in Mtb, we identified proteins whose involvements in these pathways were not previously suspected. Furthermore, the subcellular localizations of these expressed proteins provide interesting insights into the compartmentalization of these pathways, which appear to traverse from cell wall to cytoplasm. Results of this large-scale subcellular proteome profile of Mtb have confirmed and validated the computational network hypothesis that functionally related proteins work together in larger organizational structures.

Figure 1.

Subcellular fractionation of Mtb cell lysate. Differential centrifugation was utilized to obtain three subcellular compartments: the cell wall, a membrane fraction, and a cytosolic fraction. The cell wall, a membrane fractions were extensively washed to remove loosely attached and potential contaminant proteins. The separation method is illustrated in the figure.

Figure 6.

Response networks and subnetworks created by computational methods. The overview pathways were created for (A) the fatty acid degradation and (B) lipid biosynthesis. In both networks, rectangular nodes represent proteins classified as belonging to the fatty acid degradation pathway, oval nodes as belonging to the lipid biosynthesis (The Wellcome Trust Sanger Institute), and hexagonal nodes as belonging to neither. Red node proteins were identified in the cytoplasm, blue in the cell wall, and green in the membrane. Proteins found in both the cell membrane and cell wall were colored pale blue, and proteins found in both membrane and cytoplasm were colored orange. No common proteins were found in the cell wall and cytoplasm, which validates the fractionation procedure and reflects a true compartmentalization of the proteins. White node proteins were not identified. Black connections represent those identified biochemically (Marcotte et al., 1999), whereas green connections represent Rosetta Stone predictions (Marcotte, 2000; Enright and Ouzounis, 2001). The pathway involving accA3, accD3, fabD, kasA, Rv2182c, kasB, and fabD was found in both networks in A and B.

Figure 2.

Flow diagram of the Dijkstra and REA algorithms used to identify response networks. Small cap roman numbers at each subprocess refer to specific tasks discussed in the text.

Figure 3.

A Venn diagram indicating the distribution of proteins identified. Peptides identified from the analysis of each fraction were matched to 1044 nonredundant proteins. The number of proteins detected in the cytosolic fraction was 356 (173 unique), cell wall 306 (131 unique), and the membrane 705 (464 unique).

Figure 4.

The plot of variations in functional distribution of genome-encoded proteins compared with those of the proteome. The numerical labels on the pie charts correspond to the functional classes shown on the right-hand side. The sizes of the various pies of the chart show varying percentages of proteins predicted in the genome or identified in our proteome profile. Functional classes listed in Table 1 that did not vary in both the genome and proteome were grouped together and listed under “Others,” which include (I.J) Broad regulatory functions, (IV.B) IS elements, Repeated sequences, and Phage, (III.A) Transport/binding proteins, (I.G) Biosynthesis of cofactors, prosthetic groups and carriers, (II.B) Degradation of macromolecules, (I.H) Lipid Biosynthesis, (I.F) Purines, pyrimidines, nucleosides and nucleotides, (IV.A) Virulence, (III.F) Detoxification, (IV.D) Antibiotic production and resistance, (IV.J) Cyclases, (IV.E) Bacteriocin-like proteins, (IV.G) Coenzyme F420-dependent enzymes, (IV.K) Chelatases, and (I.E) Polyamine synthesis.

Figure 5.

Number of TM-helices in membrane proteins identified in the different cellular compartments as they compare to predicted domains in the genome. The different profiles were color-coded as shown above. The percentage of occurrence of membrane proteins domains with a given number of TM-helices were expressed as a percentage of the respective number of proteins found in each category. The relative number of TM proteins identified in the proteome corresponds with that of Mtb H37Rv genome, which possesses a large number of proteins with a smaller number of TM-helices, i.e., <7, and the number of proteins with >7 TM-helices decreases drastically. Also, proteins with TM-helices >10 predominate in the proteomics profile, which has the highest number of TM proteins in the membrane compartment, as expected, followed by cytosolic membrane proteins and then TM proteins found in the cell wall. SP, secreted proteins.