table of contents table of contents

A novel and comprehensive meta-analysis of the M. tuberculosis gene expression and …

Home » Biology Articles » Microbiology » Microbial Genetics » Identification of gene targets against dormant phase Mycobacterium tuberculosis infections » Results

- Identification of gene targets against dormant phase Mycobacterium tuberculosis infections

Figure 1 shows a flowchart of our approach to utilize the microarray data sets to identify putative gene targets in non-replicative M. tuberculosis. Table 1 shows the data sets collected in the first step. Since the experimental conditions in the dormancy models were quite varied (e.g. 24 h of starvation in culture media to 4 weeks in whole mice), the expression results for each gene were normalized (Fig. 1, Step 2). A zero to five scoring system was developed based upon two criteria. The first criterion was the overall relevance of the experimental conditions to persistance in the granuloma. The mouse macrophage and whole animals studies model the immediate response of M. tuberculosis to immune attack and long term survival in the host. The granuloma itself is characterized by avascularization and necrosis which have been modeled by the hypoxic and starvation conditions. The maximum score for a particular experimental dataset was adjusted based on potential relevance to the clinical occurrence of dormancy phase M. tuberculosis infections. For studies with multiple time series sampling, increasing weight was given to later time-point samples. The second criteria involved the rank order of gene expression in a particular study which allowed for cross-study comparsions (See Table 1 and Methods for details on the scoring scheme). Down-regulated expression was scored the same as up-regulated expression except that negative values were used to easily separate the two sets of results. (Some genes show significant scores in both the up-regulated and down-regulated data sets. This is not surprising considering the variation among studies in experimental situation and the number of time-points.) The knockout experiments were similarly scored by rank order of effect on growth, except that genes having no effect were scored as zero.

In Step 3 scores for each gene in each of the experimental conditions were collected into a Microsoft Access database. Reference fields were added to facilitate prioritization, such as the Refseq ID, Genbank function, Genbank note, Tuberculist classification, and KEGG and Sanger Center links. These data are available in Additional file 1.

There are two important characteristics of this meta-analysis: i) in order for a gene to score well it must be in the top quarter of highly induced genes across several experimental models of dormancy and ii) the expression levels for the highest expressing genes are attenuated. The latter item has the effect of avoiding the situation where a very large fold increase in one experiment dwarfs all other results. The first point is illustrated by the intersection of the top 400 genes (~10% of the genome) from the hypoxia, starvation, and in vivo murine models, shown in Fig. 2. The great majority of the high scoring genes come from the subset where two or three of the groups intersect. By combining the data from different models, a consensus view can reached about the particular genes and pathways most critical for survival in the dormant state.

Multi-gene trends

Figure 3 shows a comparison of the functional classes [17] of the up-regulated and down-regulated genes to the whole genome. The proportion of genes in the top 10% of up or down-regulated genes was divided by the proportion of that functional class in the entire genome and the ratio is plotted. The following sections highlight differential changes in particular functional classes shown in Fig. 3 as well as other multi-gene expression results that impacted the selection of therapeutic targets.

devR regulon

There is significant experimental evidence that the transcriptional regulator devR (also called dosR) is a key factor in the metabolic shift-down to non-replicative persistence (see Discussion). Based on the 20 nt consensus sequence of devR and microarray expression results, two studies have identified 53 genes that appear to be induced in response to devR activation [18,19]. The analysis here shows 81% of the 53 genes in the devR regulon are in the top 5% of highest scoring genes, confirming the importance of this regulatory element in dormancy under a variety of experimental conditions (see Additional file 2). Only up-regulated genes were extracted from the data sets [18,19] so it is expected that these 53 genes show higher than average up-regulation scores (see Kendall [20] and Ohno [21] for analyses that included genes down-regulated by devR). Fig. 4A shows a comparison of the expression pattern for the entire genome and the devR regulon.

While the 53 genes regulated by devR appear to play an important role in dormancy, nearly 60% of the genes do not have an annotated function. In order to gain some insight into the genes regulated by this system, we searched against INTERPRO (version 12.0) and the COGS database [22,23]. While some of the assignments are speculative, (i.e. based on partial overlap with a domain of known function [24]) several useful clues emerge. Eleven genes are involved in carbohydrate and fatty acid metabolism, and eight genes function in electron transfer. We suggest these genes reflect the biased nutrient pool and lack of oxygen, requiring altered pathways for biosynthesis and to generate oxidizable metabolites and utilization of other terminal electron acceptors. The occurrence of four transporters also indicates a limited nutrient pool. Among the other genes of the devR regulon, only two genes alter translation compared with possibly seven transcriptional regulators. This makes the devR regulon even more significant when one considers the cascading signal that is produced. These and the other regulators (36 of the 190 total are in the top 10%) show that M. tuberculosis has to make global changes to achieve a dormant state. There are several genes involved in nucleotide biosynthesis despite the fact that the genome has been replicated prior to entry into the non-replicative state [25]. This can be rationalized by the need to make repairs to maintain the integrity of the genome over decades [26]. The signal from the expression experiments for the universal stress proteins (USPs) is very strong: six of the eight USPs are part of the devR regulon, and five of those are in the top 5% of up-regulated genes.

Only six of the 53 genes have an effect on growth in the gene disruption experiments (TraSH), supporting the idea that their main role is in shifting to and maintaining the non-replicative state. DevS, a regulator of devR (see below) inhibits growth in vitro [13] and significantly in the mouse macrophage model [14], consistent with the response to nitric oxide [18], Disruption of one of the USPs inhibits in vitro growth [13]; the others had no effect. The gene with the highest growth inhibition score (9.4) was Rv2004c. This gene has a partial overlap and low similarity to gluconate kinase, but is highly similar to COG2187, a conserved bacterial domain of unknown function.

C2 metabolism

Fig. 3 shows modest changes in the proportion of genes involved in intermediary metabolism and respiration. We interpret this to mean M. tuberculosis has kept a portion of its metabolic repertoire intact to adapt to hypoxia and a biased nutrient supply. Up-regulation of genes in C2 metabolism, especially enzymes that maintain redox balance, utilize alternate terminal electron acceptors, or handle the increased two-carbon flux point to these proteins as particularly important in the metabolic alterations made for survival in the hypoxic granuloma.

Downregulation of ATP synthesis

Consistent with previous work and analysis [9,27] the data here shows a strong down-regulation of the F1-FO ATP synthase (see Additional file 3), a likely consequence of hypoxia and the utilization of other terminal electron acceptors. It thus seems probable that ATP is a scarce resource in the non-replicating cell.

Down-regulation of ribosomal proteins

Hu et al. [28] showed a 98% decrease in protein synthesis using 35S-met pulse labeling experiments with microaerophilic cultures (50 days in unstirred, screwcap vials) similar to the Wayne model [8]. The transcriptional experiments collected here are consistent with this result. Over half of the 55 30S and 50S ribosomal-protein genes encoded by the M. tuberculosis genome are down-regulated. Fig. 3 also illustrates that information pathways (including transcription and translation) show significant down-regulation.

Isoprene biosynthesis

Fig. 5 shows the isopentenyl-pyrophosphate biosynthetic pathway in M. tuberculosis. Of all the metabolic pathways we examined, this pathway showed the most consistent up-regulation of genes across the entire pathway. One of the uses of isopentenyl-pyrophosphate is the biosynthesis of decaprenyl phosphate, which is needed for cell wall construction (see Discussion). Consistent with this role is the result that insertion into dxs1, ispD, ispE and ispF show inhibition in the in vitro growth experiments [13]. Significant synthesis of cell wall components is also needed to survive the multiple stresses inside the macrophage phagosome. Nearly all of genes in the pathway are up-regulated in the mouse macrophage model [29]. Also contributing to the high expression scores are the later time points in the long terms experiments [30,31]. This may reflect the continuing need to maintain membrane integrity during long term survival.

rating: 1.00 from 1 votes | updated on: 28 Aug 2007 | views: 10523 |

Rate article: