The Msm transcriptome

A gene-expression profile collected for the whole genome for cells grown in reference condition indicates that nearly the entire genome was probed in the array. The reference medium is composed of Middlebrook 7H9 broth, supplemented with glucose, glycerol and Tween 80 and reflects a standard wild type condition. We term this as the `reference condition hereafter. The expression patterns for the 2 biological replicates were seen to be highly similar (Figure 2) with a very high correlation coefficient (R = 0.99). Hence, average gene expression was calculated for the samples and the value has been used for other analyses [25]. Frequency distributions of normalized gene expression in the replicate arrays showing similar pattern in both, reflects a normal distribution. About 5018, 3278, 1597 and 676 genes out of 6761 genes probed using the microarray chip showed values higher than 25^th, 50^th, 75^th and 90^th percentile expression respectively (Additional file 2).

Identifying feasible metabolic pathways

In order to identify metabolic pathways active in Msm in log phase cultures in the reference condition, we map inferred enzyme abundances from gene expression values of individual genes, for all pathways in the organism listed in standard databases, KEGG and BioCyc [26],[27]. For a pathway to be active, enzymes in it must be expressed in detectable quantities. Although gene-expression does not always directly correlate with protein abundances, transcription data is clearly suggestive of whether or not a protein is present in detectable quantities. Moderate correlation between expression levels and protein abundances has been reported for bacterial systems [25],[28]. 338 pathways are identified for Msm that combines knowledge of experimentally known pathways from literature along with those inferred from genome sequence analysis. Genes corresponding to enzymes in expected pathways including central carbon metabolism, amino acid biosynthesis, purine and pyrimidine biosynthesis, fatty acid metabolism, mycolic acid biosynthesis are all expressed, as expected. Figure 3a and b shows gene expression pattern corresponding to enzymes in some pathways (data for all 338 pathways is given in Additional file 2), which illustrates that many pathways including those of central carbon metabolism, as expected, appear active owing to expression of all required genes. However expression levels vary from low to high, which is quite understandable owing to their individual biochemical properties. In all, 257 pathways can be considered to be active in the condition studied (for example, the first and second row in Figure 3b). There are about 14 pathways in which the genes show no expression (for example, last row in Figure 3b), and about 57 pathways where few genes in them are expressed whereas 75 pathways had most of the genes expressed. The latter have implications of ease of adaptability (discussed in a later section).

Another interesting feature that emerges from this analysis is the identification of the active enzyme(s) from the set of duplicates available for a given reaction. We analysed the expression patterns of 24 such sets of duplicate genes in terms of their contribution to their respective pathways. The trend that we observe indicates that in most cases only one of the possible alternatives is expressed (above median levels), while others are not, reflecting that there is minimisation of cellular expenditure in expressing redundant enzymes. In very few cases, more than one gene at a given step are simultaneously expressed. Figure 3c summarises our observations. Enzymes such as glucose-6-phosphate isomerase, 6-phosphofructokinase, fructose-bisphosphate aldolase, phosphoglycerate mutase, pyruvate kinase are encoded by more than 1 gene. MSMEG_3086, MSMEG_6785 code for triose phosphate isomerase enzyme. A multiple sequence alignment shown in Figure 3d indeed indicates that they are similar to each other [29]. It is interesting to observe that, of these two enzymes, only MSMEG_3086 is expressed. They are located 3674434bp away from each other at positions in the chromosome. Similarly, other sets of paralogues are also located far apart from each other in the genome, indicating different transcriptional regulation. This analysis helps in associating specific genes to individual pathways, which becomes necessary for systems level modeling of metabolism, understanding of genomic deletions and any such genotype to phenotype associations. More examples of enzymes present in central carbon metabolism are shown in Additional file 3.

Transporters for uptake of nutrients

For a compound to serve as a nutrient source, it needs to be taken up by the cell. Such uptake takes place with the help of specific transporters. We identify transporters from the genome sequence and then feasibility of their activity through gene expression data. About 282 transporters were annotated by our analysis and amongst them we found about 60 to be expressed in the reference condition (glucose as carbon source) (shown in Additional file 9). It is known that a gene cluster comprising MSMEG_2116 to MSMEG_2120 forms a part of the glucose-sucrose subfamily in phosphotransferase system (PTS) [11]. The expression of this cluster seems to be lower but these are known to be constitutive as compared to other transporters. This PTS also comprises of trehalose, GlcNAc (N-acetylglucosamine), and dihydroxyacetone (MSMEG_2121 to MSMEG_2124) permeases which are expressed. The transporters for fructose such as MSMEG_6802, MSMEG_6803 and MSMEG_6804 seem to be expressed in the reference condition itself. Msm has glucose-6-phosphate isomerase (MSMEG_5541) for its utilization. Fructose is also known to have another mechanism of utilization via the expression of fructose-specific PTS composed of EI (ptsI), HPr (ptsH), and IIABCFru (fruA) (MSMEG_0084 to MSMEG_0088). However the expression here is lower except in the first locus. This cluster is known to be inducible in the presence of fructose as the sole carbon source. Indeed, high growth is observed in the PM plate, with fructose as the carbon source. Additional file 10 lists the possible transporters as deduced from the genome sequence and highlights those among them that are expressed under reference nutrition conditions. Transporters for glucose, xylose are seen to be expressed, providing first level explanation for utilization of these compounds as carbon or nitrogen sources. It has been reported that Msm can utilize different sugars indicating activation of various transporters and hence also changes in gene expression levels [11],[39]-[41].

Connecting nutrient sources to metabolic pathways

Next, we study, if a given source compound can be mapped onto specific pathways in Msm, through which it can enter metabolism [18]. About 135 of these compounds are direct metabolites in the network and hence growth patterns with them are easily interpreted. Several more compounds can be linked with a metabolite in the network with one or few steps. In such cases, we study if the enzymes corresponding to their conversion can be detected in the genome. Additional file 4 lists these cases. One example is D-Mannitol, which is known to get converted into D-fructose then to fructose 6-phosphate, thus entering glycolysis [42]. A transporter for this can be traced from the genome sequence (MSMEG_5574). It is not expressed highly in the reference condition, but perhaps gets induced when mannitol is the sole carbon source. Similar behaviour is observed for trehalose, sorbitol and D-saccharic acid sources [11]. There are many lines of evidence from individual molecular biology studies to support the functional roles of these molecules [43]-[48]. Put together, they explain why these compounds serve as carbon or nitrogen sources that promote bacterial growth.

Another example is the conversion of serine into many central carbon metabolites through the glycolysis pathway and then to glycine and cysteine thus supporting growth [34]. The central carbon metabolism in Msm is represented in Additional file 3. Utilization of the range of carbon sources shows the repertoire of possibilities for metabolic pathways in the bacterium. Glycerol, arabinose, mannose, D-glucose and many other polyols, pentoses, hexose and also complex sugars enhance growth as supported by literature [39]-[41]. The alternate carbon sources such as L-proline, rhamnose, xylose and others are also being utilized for growth in Msm indicating these can induce their uptake and successful utilization. Glycerol can be taken up by a facilitator (MSMEG_6758) and used by the enzyme glycerol kinase (MSMEG_6759 shown to be expressed abundantly) to form glycerol 3-phosphate which can then enter central carbon metabolism. The observation about absorption and utilization of maltose is also in line with other experimental evidences, showing very low or retarded growth. Galactose and lactose show only moderate growth, consistent with the observation that the corresponding enzymes and transporters show poor expression values [11],[39]-[41].

It is not clear whether there are any transporters for utilisation of trehalose from the external medium in Msm. Nevertheless, it seems to be enabling growth in the bacterium. It is possible that it can be involved in central carbon metabolism as well as be a component of the cell wall in the form of conjugates of mycolate, such as trehalose dimycolates and trehalose monomycolates [44],[46]. Many of the TCA intermediates such as succinic acid, citric acid seem to promote growth. This is again consistent with the observation, that many of the central carbon metabolism genes are constitutively expressed in Msm. We also observe acetate and oleic acid (derived from Tween 80) being utilized for growth. This observation is consistent with known biochemical studies that glyoxylate shunt is prominent for anaplerosis in the bacterium allowing the utilization of acetate or fatty acids as the sole carbon sources while it allows the regeneration of the four-carbon malate from glyoxylate and acetyl-CoA for biosynthetic processes. The shunt can also replenish amino acids such as glycine and serine [34].

Amino acids such as L-Proline, L-Alanine and dipeptides such as L-Alanyl-Glycine seem to promote growth as nitrogen sources [37],[49]-[51]. All other amino acids tested are also able to support growth either highly or moderately, indicating the ability of Msm to adapt to a wide variety of nitrogen sources and supplements. The genes involved in purine salvage pathways seem to be moderately expressed in Msm. Adenosine as a sole carbon source does not support growth, consistent with earlier suggestions in literature as well [31]. Examination of the gene expression values of enzymes involved in a pathway that salvages adenosine, indicates that the pathway is infeasible since enzymes adenosine deaminase, adenosine kinase and adenine phosphoribosyltransferase are virtually non-expressed under the conditions studied. However, when adenosine is supplied as a nitrogen source along with glucose as the carbon source, small extent of utilization is observed. An enzyme unique to mycobacteria, 5-methylthioadenosine phosphorylase (MSMEG_0990), that converts adenosine to adenine and alpha-D-ribose-phosphate is moderately expressed, perhaps presenting the only feasible way for adenosine utilization [52]. Thus the low activity of purine salvage pathways makes de novo biosynthesis of purine nucleotides highly essential for the survival of the organism, presenting targets for antimycobacterial drugs [53]-[56]. In fact, analog-based inhibitions of the de novo biosynthesis pathway enzymes are already under consideration as anti-tubercular drugs. Guanosine can be efficiently used by Msm as the same enzyme (MSMEG_0990) can cleave inosine and guanosine as well [52]. Overall, the differences in growth patterns under different conditions are explained by (a) presence or absence of a transporter for nutrient uptake, (b) presence and the expression level of the utilizing enzymes.

Microarray data of Mtb shows about one-fourth of the genes are consistently expressed under standard nutrition conditions in in vitro cultures [57]. Phenotypic microarray studies have been reported for Mtb, using a similar Biolog experimental setup [31],[32]. Comparison of the growth patterns in Msm and Mtb, as observed from phenotypic microarray experiments, reveals that the two species show similar growth behaviour in most cases. This implies similar metabolic flow for most of the studied probes (Table 2). Exceptions to this are compounds D-malate, D-mannose, N-acetyl glucosamine, propoanoate, allantoin, L-aspartic acid and L-threonine which serve as nutrients to Msm but not to Mtb, while D-serine is the only compound that serves as a nutrient to Mtb but not to Msm (Table 2). Thus, overall, Msm can utilise most of the carbon sources and nutritional supplements as compared to Mtb and other mycobacterial species. Tween can be used as a sole carbon source in both but unlike in Mtb, it cannot be utilised in combination with glucose in Msm. Thus, it can be seen that 31 nutrient sources are common and 9 are unique between Msm and Mtb.

a) Strain and culture condition

M. smegmatis MC² 155 wildtype culture was grown in Middlebrook 7H9 media until it reached 0.2-0.3 O.D₆₀₀. Once the O.D was reached, 20ml of the culture was pelleted down, and the supernatant was discarded. The pellet was resuspended in 100?l of 1 PBS. The pellet was snap freezed in liquid nitrogen and stored in -80C until RNA extraction was carried out.

b) RNA extraction

RNA extraction was done using Qiagens RNeasy minikit (Cat#74104). The RNA quality was checked using Bioanalyzer. Labelling was done using Agilents Quick-Amp labeling Kit. Random hexamer method of labeling was done followed by T7 promoter based-linear amplification to generate labeled complementary RNA (One-Color Microarray-Based Gene Expression Analysis). Hybridization was performed using Agilents In situ Hybridzation kit 51885242. Chips used for microarray were customized for M. smegmatis MC² 155815k Array AMADID: 020791 (Genotypic Technology, Bangalore, India).

c) Transcriptome data analysis

The raw data obtained from experiments have been normalized using GeneSpring GX 12.6.1 software. Intra-array normalization deals with variability within a single array. In intra-array normalization, gProcessed signal (dye normalized background subtracted signal intensity) is log transformed and then for each of the array elements, the 75^th percentile value is calculated separately. In each sample the log transformed intensity values for each probe is subtracted by the calculated 75^th percentile value of the respective array and expression values are obtained. Similarly 50^th and 25^th percentile normalization was calculated for the dataset.

d) Clustering

Hierarchical clustering of the normalized data was performed using GeneSpring GX 12.6.1 software. Pearson correlation coefficient to measure similarity between expression profiles and average linkage method was used for clustering genes.

a) Bacterial strains, growth conditions and chemicals

Mycobacterium smegmatis MC² 155 were cultivated at 37C in Middlebrook 7H9 broth with 0.2% (vol/vol) glycerol and 0.1% (wt/vol) Tween 80 or on Middlebrook media 7H10 agar plates supplemented with 0.2% (vol/vol) glycerol and supplemented with OAD (oleic acid, albumin, dextrose).

b) PM measurements

Phenotype microarray experiments were carried out following standard Biolog Inc. (http://www.biolog.com/) protocols as provided by the supplier. To prepare the inoculum for Phenotypic microarray plates (PM01 for carbon source, PM03 for nitrogen source, PM05 for other nutrient supplement), bacteria colonies were grown in Middlebrook 7H10 medium containing 10% (v/v) albumin dextrose (AD) enrichment and 0.05% (v/v) Tween 80. Bacteria were harvested at 48hrs. The M. smegmatis strains were re-suspended in inoculating fluid so as to have 81% transmittance. The Biolog plates PM03 and PM05 contain dextrose as a carbon source in the PM additive.

PM plates were inoculated with 100?l of the mixture made up with the following volumes per plate: Middlebrook 7H9 broth at 1.2 (10ml), Dye mix G at 100 (0.12ml), PM additive appropriate to the plate at 12 (1ml) and bacteria in the media at 13.64 (0.88ml). For each plate, the final volume of mixture was 12ml. After plate inoculation, the plates were transferred to an OmniLog (Biolog, Inc.) incubator and incubated at 37C for 4days and monitored for color change due to dye reduction in the wells. To have biological replicates, separate inocula were used in the experiment.

c) Data processing and analysis

Data were analysed initially with OmniLog-PM software for gathering the kinetic values or respiration rates. Then for further analysis, data aggregation, discretization and clustering of the biological replicates in each PM plate types were carried out using OPM package available in R. MATLAB R2011b Toolboxes were used for clustering and correlation analysis across PM plates.