Evaluation of a nutrient starvation model of Mycobacterium tuberculosis persistence by gene and protein expression profiling



The search for new TB drugs that rapidly and effectively sterilize the tissues and are thus able to shorten the duration of chemotherapy from the current 6 months has been hampered by a lack of understanding of the metabolism of the bacterium when in a ‘persistent’ or latent form. Little is known about the condition in which the bacilli survive, although laboratory models have shown that Mycobacterium tuberculosis can exist in a non-growing, drug-resistant state that may mimic persistence in vivo. Using nutrient starvation, we have established a model in which M. tuberculosis arrests growth, decreases its respiration rate and is resistant to isoniazid, rifampicin and metronidazole. We have used microarray and proteome analysis to investigate the response of M. tuberculosis to nutrient starvation. Proteome analysis of 6-week-starved cultures revealed the induction of several proteins. Microarray analysis enabled us to monitor gene expression during adaptation to nutrient starvation and confirmed the changes seen at the protein level. This has provided evidence for slowdown of the transcription apparatus, energy metabolism, lipid biosynthesis and cell division in addition to induction of the stringent response and several other genes that may play a role in maintaining long-term survival within the host. Thus, we have generated a model with which we can search for agents active against persistent M. tuberculosis and revealed a number of potential targets expressed under these conditions.


It takes at least 6 months of antimycobacterial chemo-therapy to cure a patient of tuberculosis because the drugs available today are poorly active against an ill-defined population of ‘persistent’Mycobacterium tuberculosis bacilli. New drugs with potent sterilizing ability are urgently needed in order to shorten the duration of drug treatment and reduce the risk of relapse, whereas agents active against latent TB would bring about elimination of the disease in low-incidence countries (O’Brien and Nunn, 2001). However, we understand little of the physiological state of persistent M. tuberculosis, and few well-validated targets for intervention are available. In vitro models that mimic the persistent state are required for the identification and testing of novel agents.

Controversy exists as to the location of M. tuberculosis organisms in the absence of clinical disease. The most common assumption is that the viable bacilli reside within fibrotic granulomatous lesions within the lung and are maintained in a dormant state through conditions such as low oxygen tension and nutrient limitation (Wayne, 1994). However, M. tuberculosis DNA has recently been detected within intracellular compartments in lung tissue from persons with no evidence of ongoing or previous disease (Hernandez-Pando et al., 2000). In this case, other pressures, such as the immune system, restrict the development of active disease.

Models using conditions of reduced oxygen, nutrient deprivation and stationary phase survival have demonstrated that M. tuberculosis is able to survive for extended periods in a non-growing state. Wayne chose to mimic the ‘dormant’ state in vitro through oxygen limitation (Wayne, 1994; Wayne and Hayes, 1996). M. tuberculosis was able to adapt to and survive anaerobiosis. The bacilli were in a non-replicative state and exhibited resistance to the drugs isoniazid and rifampicin, yet became sensitive to metronidazole (Wayne and Sramek, 1994). Others have used the Wayne model to identify genes and proteins induced in M. tuberculosis and Mycobacterium bovis BCG during oxygen limitation (Cunningham and Spreadbury, 1998; Hu et al., 1999; Hutter and Dick, 1999; Boon et al., 2001).

There is also evidence to suggest that persistent bacilli in lung lesions suffer nutrient deprivation. M. tuberculosis isolated from lung lesions has been found to display altered morphology and staining properties compared with bacilli grown in vitro (Nyka, 1974). M. tuberculosis cultures starved in distilled water were found to display these same properties but regained acid fastness and started growing when added to nutrient-rich medium, even after a 2 year starvation period (Nyka, 1974). Loebel et al. (1933a,b) used a nutrient starvation model in order to investigate the effect of nutrients predicted to be available in a granuloma on the metabolism of M. tuberculosis. Cultures were transferred from nutrient-rich medium into phosphate-buffered saline (PBS) and the respiration rate measured using a manometer. Nutrient starvation resulted in a gradual shutdown of respiration to minimal levels, but bacilli remained viable and were able to recover when returned to rich medium (Loebel et al., 1933b). Mycobacterium smegmatis starved of carbon, nitrogen or phosphorous has been shown to remain viable for over 650 days after an initial 2- to 3-log drop in colony-forming units (cfu) and displayed increased stress resist-ance, increased mRNA stability and an overall decrease in protein synthesis (Smeulders et al., 1999).

In an effort to design a simple model in which to test novel antimycobacterial drugs active against persistent bacteria, we have developed a nutrient starvation model based on Loebel’s work (Loebel et al., 1933b), in which M. tuberculosis is in a state of little or no replication, exhibits a low respiration rate, yet maintains long-term viability. In addition, cultures are resistant to both isoniazid and rifampicin, yet not sensitive to metronidazole. Through proteome analysis, we have been able to identify proteins induced under these conditions and have then gone on to characterize gene expression patterns occurring as the bacilli adapt to starvation using DNA microarrays.

Results and discussion

Bacterial viability and oxygen consumption during starvation

Cultures grown for 7 days in nutrient-rich media were pelleted, washed twice with PBS and then resuspended in PBS and left standing at 37°C in sealed bottles. The viability of M. tuberculosis during starvation in PBS was determined by counting cfus from triplicate cultures at five time points over a 6 week period. No loss of viability occurred during this time, with cfu levels remaining constant at ≈ 107 cfu ml–1 at all time points. The addition of methylene blue dye to cultures in nutrient-rich broth showed complete dye decolourization and, hence, oxygen depletion after ≈ 9 days. In contrast, methylene blue dye in the nutrient-starved cultures did not decolourize but remained the same colour as the control solution containing no bacteria. It seems, therefore, that the starved cells significantly decreased their respiration rate, yet remained viable in the presence of oxygen. These findings are in agreement with the observations of Loebel et al. (1933b), who found that the respiration rate of nutrient-starved cultures declined rapidly over the first 96 h of starvation. This is a key difference between the nutrient starvation model described here and the Wayne model in which M. tuberculosis enters a dormant state as a result of oxygen depletion (Wayne, 1994).

Drug treatment of starved cultures

Six-week-starved and log-phase cultures were treated for 7 days with the drugs rifampicin, isoniazid or metronidazole in duplicate at concentrations of 1 and 10 μg ml–1. Figure 1 shows the effect of these treatments on culture viability compared with cultures incubated without drugs. As expected, the growing log-phase cultures are sensitive to both rifampicin and isoniazid at each concentration, but are resistant to metronidazole (Fig. 1A). Colony-forming unit counts of the log-phase cultures are reduced by 4.5 and 6.0 logs for rifampicin and isoniazid, respectively, at the 1 μg ml–1 concentration compared with the control. At 10 μg ml–1, cfus are reduced by 4.5 and 5.5 logs for rifampicin and isoniazid, respectively, compared with the control. In contrast, the nutrient-starved cultures are resistant to both rifampicin and isoniazid at 1 μg ml–1 with no change in cfus relative to the control (Fig. 1B). At 10 μg ml–1, starved cultures are still resistant to isoniazid with only a 1.5 log drop in cfus relative to the control. However, resistance of the starved cultures to rifampicin at this higher concentration of drug, which equates to 100 times the minimum inhibitory concentration (MIC), is less marked with a 3.5 log drop in cfus rela-tive to the control (Fig. 1B). Metronidazole had no effect on the viability of the starved cultures. Again, this is in contrast to the Wayne model, in which cultures that have shifted into a dormant state as a result of oxygen limitation become sensitive to metronidazole (Wayne and Sramek, 1994). As metronidazole is known to require reduction under anaerobic conditions for activity, these results confirm that the starved cultures are not in an anaerobic state. Metronidazole has been found to be in-effective in preventing reactivation in the Cornell mouse model; however, it does show some activity against bacteria during the very late chronic stage of infection (Brooks et al., 1999).

Figure 1.

Resistance of starved cultures to isoniazid, rifampicin and metronidazole. The effect of drug treatment on the viability of log-phase and 6-week-starved cultures was investigated. Cultures were treated in duplicate with rifampicin (RIF), isoniazid (INH) or metronidazole (MTZ) at concentrations of 1 μg ml–1 (dark grey bars) and 10 μg ml–1 (light grey bars) for 7 days. Duplicate control samples of both starved and log-phase M. tuberculosis were incubated without drug for 7 days. After this time, bacteria were pelleted, resuspended in media without drug and viability assessed by cfu counts for (A) log-phase cultures and (B) 6-week-starved cultures.

Two-dimensional gel electrophoresis

In order to investigate proteins expressed by nutrient-starved cultures showing resistance to antibiotics, cell lysate proteins of M. tuberculosis starved in PBS for 6 weeks were analysed by two-dimensional (2D) gel electrophoresis, and protein spot patterns were compared with those of log-phase-grown cultures (Fig. 2). We chose to study protein profiles after 6 weeks because Loebel’s work demonstrated that, by this time point, the respiration rate of M. tuberculosis had reached minimal levels (Loebel et al., 1933b), and they are therefore likely to represent a homogeneous cell population. This was also the time point used to test drug susceptibility. Cultures were grown in duplicate and pooled for 2D electrophoretic analysis. Duplicate gels were run for each sample over two pH ranges (pH 4–7 and pH 3–10), and proteins were visualized by silver staining. There was no obvious evidence of protein degradation in the starved cultures. Protein spots differing between the control and starved cultures were excised from the gel and identified by either matrix-assisted laser desorption-ionization (MALDI) mass spectrometry (MS) or, if this proved unsuccessful, by online liquid chromatography tandem mass spe-ctrometry (LC/MS/MS). The protein spot identities are listed in Table 1. Only one or two peptides were identified for some proteins using LC/MS/MS, probably because of the low level of protein in the 2D gel spots analysed. However, as tandem mass spectrometry enables sequence data for peptides to be obtained, these data were sufficient to identify the corresponding protein unambiguously.

Figure 2.

Two-dimensional gel analysis of proteins of M. tuberculosis during (A) log-phase growth compared with (B) starvation in PBS for 6 weeks. Proteins were separated by isoelectric focusing on pH 4–7 IPG strips in the first dimension and 12% SDS–PAGE in the second dimension. Proteins were visualized by silver staining. Open circles and open squares represent protein spots increased or decreased under each condition respectively. Proteins spots identified are numbered and their identities listed in Table 1.

Table 1. Summary of protein spot differences identified.
Spot no.ORFGene
Gene productIdentification
No. of peptides
coverage (%)
1Rv2462c tig Chaperone protein, similar to trigger factorMALDI1655.6
2Rv1860 modD Precursor of Apa (45 kDa antigen)LC/MS/MS210.8
3Rv1860 modD Precursor of Apa (45 kDa antigen)LC/MS/MS310.5
4Rv1860 modD Precursor of Apa (45 kDa antigen)LC/MS/MS29.2
5Rv1860 modD Precursor of Apa (45 kDa antigen)LC/MS/MS29.2
6Rv0351 grpE GrpE, stimulates DnaK ATPase activityMALDI733.2
7Rv1980c mpt64 Secreted immunogenic protein Mpt70LC/MS/MS632.9
8Rv255724.3 kDa conserved hypothetical proteinLC/MS/MS745.5
9Rv255724.3 kDa conserved hypothetical proteinMALDI1162.0
10Rv255825.7 kDa conserved hypothetical proteinMALDI835.6
11Rv2031c hspX 16 kDa antigen, α-crystallin homologueLC/MS/MS216.7
12Rv2031c hspX 16 kDa antigen, α-crystallin homologueLC/MS/MS111.8
13Rv2031c hspX 16 kDa antigen, α-crystallin homologueLC/MS/MS216.7

Seven protein spots were seen to decrease in intensity as a result of starvation (spots 1–7; Fig. 2A; Table 1). Tig is a chaperone-like protein homologous to the Bacillus subtilis trigger factor, thought to be involved in protein export. The 45 kDa antigen and MPT64 are both found in culture filtrates of M. tuberculosis (Sonnenberg and Belisle, 1997). The 45 kDa antigen is a glycoprotein (Dobos et al., 1996), which was identified as being recognized by antibodies and producing a delayed-type hypersensitivity (DTH) response in immunized guinea pigs (Romain et al., 1993). MPT64, which is deleted from a number of BCG strains, is recognized by the immune systems of the majority of TB patients and their contacts (Roche et al., 1994; 1996). GrpE functions together with DnaJ as a co-chaperone for DnaK (Bukau and Horwich, 1998). The disappearance of MPT64 and the 45 kDa antigen from gels of the starved cultures could be explained by their export into the extracellular environment and, hence, they were not present within cell lysates, whereas spots identified on gels of the log-phase cultures may represent proteins located in the cell wall before export.

Six protein spots (spots 8–13, Fig. 2B) were seen to increase in intensity in the starved cultures. Spots 8 and 9 were both identified as the 24.3 kDa hypothetical protein (Rv2557), and spot 10 was identified as the 25.7 kDa hypothetical protein (Rv2558). The genes encoding these proteins are adjacent in the genome. They bear 69% identity to each other at the amino acid level, but their function is unknown. Direct homologues of both are present in the M. bovis genome, but just one copy is present in the M. smegmatis and Mycobacterium avium genomes, and there is a pseudogene in Mycobacterium leprae. Sequence-based analyses identified no known functional motifs or domains. Three further protein spots (spots 11–13) that were seen to increase in intensity on gels of the starved cultures were identified as the 16 kDa antigen α-crystallin homologue. Expression of the α-crystallin protein has been shown to increase during the transition from log-phase to stationary-phase growth and under conditions of hypoxia and is thought to function as a molecular chaperone (Yuan et al., 1996; Boon et al., 2001). Other major protein spots identified as α-crystallin are marked on Fig. 2 and are present on gels of both starved and log-phase cultures. α-Crystallin spots showing increased intensity on gels of the starved cultures are therefore species with differing pI and molecular weights compared with the most abundant forms. Different species of this protein at lower molecular weights and differing pIs have been observed previously (Sonnenberg and Belisle, 1997; Betts et al., 2000) and could result from degradation, post-translational processing or post-translational modification, such as phosphorylation.

Microarray analysis

In a more sensitive and global approach to the analysis of the adaptation of M. tuberculosis to nutrient starvation, we used glass slide microarray analysis to monitor the initial transcriptional response. Relative gene expression profiles of M. tuberculosis starved in PBS for 4 h, 24 h and 96 h compared with PBS-washed, 7-day-old, log-phase cultures were determined. We chose these time points in order to track gene expression changes as the cultures responded to and adapted to the starvation conditions and also to ensure that RNA isolated was both abundant and of good quality. The time points chosen reflected the work of Loebel et al. (1933b), in which the decrease in respiration rate of starved M. tuberculosis declined most rapidly over the first 96 h. RNA was isolated from triplicate starved cultures at each time point. RNA isolated from triplicate, PBS-washed, log-phase cultures (t = 0) was subsequently pooled and used as the reference to which each replicate of the later time points was compared by co-hybridization to the array. Two hybridizations per replicate sample were performed for the 4 h time point, giving a total of six hybridizations. Each replicate of the 24 h and 96 h time points was hybridized once giving a total of three hybridizations for each of these two time points. The Cy5/Cy3 fluorescence ratios calculated for each open reading frame (ORF) represented on the array therefore enabled relative changes in gene expression across the time course as a result of starvation to be monitored.

Whole-genome perspective

ANOVA was used to determine whether the expression ratio at each time point was significantly different from 1. Genes were considered to be significantly differentially expressed compared with the log-phase control if they displayed at least a twofold induction or repression and a P-value <0.001. Genes showing altered expression ratios, categorized according to function, are displayed in Table 2. The relatively large number of differentially expressed genes indicates that the nutrient starvation conditions have induced a global shift in gene expression. There is a gradual increase in the number of genes both up- and downregulated with time. After 96 h, 323 genes are expressed at significantly lower levels compared with the log-phase growth control. There is a noticeable trend towards downregulation of genes involved in amino acid biosynthesis, biosynthesis of cofactors, prosthetic groups and carriers, DNA replication, repair, recombination and restriction/modification, energy metabolism, lipid biosynthesis, translation and post-translational modification and virulence. Functional classes showing upregulation across the time course include antibiotic production and resistance, insertion sequence (IS) elements, repeated sequences and phage, nucleotide biosynthesis and metabolism, putative enzymes and regulatory function. The cell envelope, cell processes, conserved hypotheticals, unknowns, PE and PPE families and RNA synthesis, RNA modification and DNA transcription functional classes show subsets of genes both up- and downregulated. Whereas the repressed genes seem to indicate general shutdown of metabolic and non-essential processes, many of the induced genes are likely to be involved in the switch in metabolism from conditions of nutrient excess to starvation conditions and are important for maintaining long-term survival.

Table 2. Summary of gene numbers, according to functional classification, up- and downregulated during the starvation time course.
Functional classificationaNo. of genesb
4 h24 h96 h
  • a. According to Cole et al. (1998).

  • b.

    Number of genes showing significant expression ratios (twofold induction or repression and a P-value < 0.001).

Whole genome39243649170211252301279323
Amino acid biosynthesis969218214113
Antibiotic production and resistance1715303040
Biosynthesis of cofactors, prosthetic groups and carriers11711027213312
Cell envelope360337211916302334
Cell processes20620182118201920
Central intermediary metabolism4544324514
Conserved hypotheticals915845404956555953
DNA replication, repair, recombination and restriction/modification73724208210
Energy metabolism29228151922421543
Fatty acid degradation11911528310711
IS elements, repeated sequences and phage13512546140140
Lipid biosynthesis65635329310
Macromolecule degradation8783332676
Nucleotide biosynthesis and metabolism6055105132
PE and PPE families1671193748610
Polyketide and non-ribosomal peptide synthesis4140361844
Polysaccharide (cytoplasmic) metabolism88000000
Putative enzymes11210592125157
Regulatory function18717761616102313
RNA synthesis, RNA modification and DNA transcription3231144466
Small molecule degradation4439313121
Translation and post-translational modification10293410418130

Functional classes showing significant proportions of differentially expressed genes and allowing some insight into the physiological and regulatory response of M. tuberculosis to starvation conditions were examined in more detail as described below and displayed in Table 3. We have focused on the changes that have taken place by 96 h, and the following observations, unless stated otherwise, are based on this time point. See the Supplementary material for the full data set for all time points.

Table 3. Summary of genes differentially expressed at 96 h discussed in the text.
ORFGene nameGene productFold change in expressiona
  • a.

    Average fold change in expression (starved/log-phase cultures) resulting from three independent hybridizations. Negative numbers indicate downregulation.

  • b. Functional class, according to Cole et al. (1998).

Energy metabolismb
Rv1098c fum Fumarase–2.81
Rv1304 atpB ATP synthase chain–5.18
Rv1305 atpE ATP synthase c chain–4.98
Rv1306 atpF ATP synthase b chain–8.00
Rv1307 atpH ATP synthase chain–3.45
Rv1308 atpA ATP synthase chain–5.75
Rv1309 atpG ATP synthase chain–8.13
Rv1310 atpD ATP synthase chain–4.65
Rv1436 gap Glyceraldehyde 3-phosphate dehydrogenase–2.47
Rv1438 tpi Triosephosphate isomerase–3.16
Rv1475c acn Aconitate hydratase–4.74
Rv1552 frdA Fumarate reductase flavoprotein subunit7.50
Rv1837c glcB Malate synthase–2.79
Rv1915 aceAa Isocitrate lyase, module2.75
Rv2495c pdhC Dihydrolipoamide acetyltransferase4.85
Rv2496c pdhB Pyruvate dehydrogenase E1 component subunit5.74
Rv2497c pdhA Pyruvate dehydrogenase E1 component subunit8.47
Rv3145 nuoA NADH dehydrogenase chain A–2.12
Rv3146 nuoB NADH dehydrogenase chain B–3.72
Rv3147 nuoC NADH dehydrogenase chain C–3.69
Rv3148 nuoD NADH dehydrogenase chain D–4.65
Rv3149 nuoE NADH dehydrogenase chain E–14.08
Rv3150 nuoF NADH dehydrogenase chain F–3.10
Rv3151 nuoG NADH dehydrogenase chain G–3.79
Rv3152 nuoH NADH dehydrogenase chain H–5.49
Rv3153 nuoI NADH dehydrogenase chain I–2.80
Rv3154 nuoJ NADH dehydrogenase chain J–4.63
Rv3155 nuoK NADH dehydrogenase chain K–5.21
Rv3156 nuoL NADH dehydrogenase chain L–2.48
Rv3157 nuoM NADH dehydrogenase chain M–3.68
Rv3339c icd1 Isocitrate dehydrogenase–2.64
Central intermediary metabolism
Rv1285 cysD ATP:sulphurylase subunit 23.04
Small molecule degradation
Rv2780 ald L-alanine dehydrogenase6.04
Lipid biosynthesis
Rv0470c umaA2 Unknown mycolic acid methyltransferase–2.11
Rv1094 desA2 Acyl-[ACP] desaturase–5.26
Rv1886c fbpB Antigen 85B, mycolyltransferase–4.46
Rv2524c fas Fatty acid synthase–3.13
Rv3229c desA3 Acyl-[ACP] desaturase–6.62
Rv3392c cmaA1 Cyclopropane mycolic acid synthase 1–2.50
Rv3804c fbpA Antigen 85A, mycolyltransferase–4.05
Polyketide and non-ribosomal peptide synthesis
Rv1660 pks10 Polyketide synthase (chalcone synthase-like)6.51
Rv2932 ppsB Phenolphthiocerol synthesis (pksC)–4.46
Rv2933 ppsC Phenolphthiocerol synthesis (pksD)–4.17
Rv2934 ppsD Phenolphthiocerol synthesis (pksE)–2.67
Rv2940c mas Mycocerosic acid synthase–3.14
Rv3825c pks2 Polyketide synthase2.90
Translation apparatus
Rv0685 tuf Elongation factor EF-Tu–2.87
Rv0700 rpsJ 30S ribosomal protein S10–2.73
Rv0701 rplC 50S ribosomal protein L3–3.01
Rv0702 rplD 50S ribosomal protein L4–5.13
Rv0703 rplW 50S ribosomal protein L23–5.43
Rv0704 rplB 50S ribosomal protein L2–2.63
Rv0705 rpsS 30S ribosomal protein S19–4.35
Rv0706 rplV 50S ribosomal protein L22–10.20
Rv0708 rplP 50S ribosomal protein L16–4.85
Rv0710 rpsQ 30S ribosomal protein S17–4.10
Rv0714 rplN 50S ribosomal protein L14–2.01
Rv0715 rplX 50S ribosomal protein L24–2.33
Rv0716 rplE 50S ribosomal protein L5–3.61
Rv0718 rpsH 30S ribosomal protein S8–2.40
Rv0719 rplF 50S ribosomal protein L6–7.87
Rv0723 rplO 50S ribosomal protein L15–5.75
Rv2534c efp Elongation factor P–2.26
Rv2992c gltS Glutamyl-tRNA synthase–5.29
Rv3336c trpS Tryptophanyl tRNa synthase–4.05
RNA synthesis, RNA modification and DNA transcription
Rv0668 rpoC β′-subunit of RNA polymerase–6.37
Rv1221 sigE ECF subfamily sigma subunit5.87
Rv2710 sigB RNA polymerase sigma factor5.48
Rv3286c sigF ECF subfamily sigma subunit2.99
Rv3287c rsbW Antisigma B factor9.14
Rv3414c sigD ECF subfamily sigma subunit2.05
Rv3457c rpoA α-subunit of RNA polymerase–2.77
Purines, pyrimidines, nucleosides and nucleotides
Rv2583c relA (p)ppGpp synthase I5.08
Rv0014c pknB Serine–threonine protein kinase–2.71
Rv0018c ppp Putative phosphoprotein phosphatase–2.90
Rv0931c pknD Serine–threonine protein kinase–4.37
Rv1027c kdpE Two-component response regulator2.20
Rv1152Transcriptional regulator (GntR family)15.29
Rv2034Transcriptional regulator (ArsR family)18.63
Rv2358Transcriptional regulator (ArsR family)4.07
Rv2359furBFerric uptake regulatory protein3.25
Rv2711 ideR Iron-dependent repressor, IdeR2.59
Rv3260c whiB2 whiB transcriptional activator homologue5.87
Rv3291cTranscriptional regulator (Lrp/AsnC family)14.57
Cell processes
Rv0930 pstA1 PstA component of phosphate uptake–5.32
Rv0933 pstB ABC transport component of phosphate uptake–2.15
Rv0934 phoS1 PstS component of phosphate uptake–3.21
Rv0936 pstA2 PstA component of phosphate uptake–2.57
Rv2397c cysA Sulphate transport ATP-binding protein3.21
Rv2398c cysW Sulphate transport system permease protein6.23
Rv2399c cysT Sulphate transport system permease protein4.35
Rv2400c subI Sulphate binding precursor4.56
Rv2462c tig Chaperone protein, similar to trigger factor–5.35
Rv3663c dppD Probable ABC transporter2.00
Rv3665c dppB Probable peptide transport system permease2.55
Rv3917c parA Chromosome partitioning; DNA binding–3.68
Rv3918c parB Possibly involved in chromosome partitioning–6.21
Cell envelope
Rv0016c pbpA Penicillin-binding protein–2.90
Rv1980c mpt64 Secreted immunogenic protein Mpt64–2.18
Rv2875 mpt70 Major secreted immunogenic protein Mpt705.53
Rv3810 pirG Cell surface protein precursor (Erp protein)11.88
Antibiotic production and resistance
Rv2036Similar to lincomycin production genes5.80
Rv3290c lat Lysine-ɛ aminotransferase41.86
Unknowns and conserved hypotheticals
Rv0116cHypothetical 26.9 kDa protein17.35
Rv1284Hypothetical 18.2 kDa protein13.93
Rv2035Hypothetical 18.5 kDa protein3.01
Rv2557Hypothetical 24.3 kDa protein8.96
Rv2558Hypothetical 25.7 kDa protein16.26
Rv2660cVery hypothetical 7.6 kDa protein282.24
Rv2661cHypothetical 13.7 kDa protein26.77
Rv2662Hypothetical 9.9 kDa protein34.23
Rv2663Hypothetical 8.9 kDa protein6.73
Rv3288cHypothetical 15.2 kDa protein10.99
Rv3289c Hypothetical 13.2 kDa protein22.16

Energy metabolism

Forty-three (15%) of the 281 energy metabolism genes on the array are significantly downregulated. This includes several enzymes of the central catabolic processes, glycolysis (gap and tpi) and the tricarboxylic acid (TCA) cycle (fum, acn and icd1), indicating a general slowdown of metabolism. The glcB gene, which encodes the glyoxylate shunt enzyme malate synthase, is downregulated, although relative expression levels of the other glyoxylate shunt enzyme, isocitrate lyase (icl), do not change. The aceAa gene is, however, upregulated. Both this gene and aceAb show homology to icl but, in H37Rv, are frameshifted and thought not to be functional (Hoener Zu Bentrup et al., 1999). Isocitrate lyase is considered to be important for in vivo survival of M. tuberculosis, as icl is upregulated upon macrophage infection (Graham and Clark-Curtiss, 1999; Hoener Zu Bentrup et al., 1999), and a mutant strain lacking this gene is unable to mount a persistent infection in mice (McKinney et al., 2000).

In accordance with the reduced respiration rate seen, 14 of the 30 genes (47%) involved in aerobic respiration are downregulated, most notable of which is the NADH dehydrogenase operon (nuoA–M). This enzyme complex functions as the primary aerobic respiratory chain and produces ATP via the oxidation of NADH to NAD+ and, in doing so, replenishes the NAD+ pool. Glycine dehydrogenase activity has been detected in M. tuberculosis and M. bovis BCG cultures under anaerobic conditions in the Wayne model (Wayne and Lin, 1982), and it has been suggested that this facilitates NAD+ recycling when oxygen as a terminal electron acceptor becomes limiting. The gene responsible for this activity is unknown. In M. smegmatis grown under similar conditions, alanine de-hydrogenase activity has been detected, which may also function to regenerate NAD+ through the conversion of pyruvate to alanine (Hutter and Dick, 1998). In M. tuberculosis, this enzyme is encoded by the ald gene, which is significantly upregulated here. It is therefore possible that the ald gene product functions to recycle NAD+ under conditions of nutrient or oxygen limitation in M. tuberculosis. Additionally, seven of the eight genes that form the ATP synthase enzyme complex (atpA–H) are downregulated [the atpC polymerase chain reaction (PCR) product was not present on the array].

The pdhABC genes are upregulated. These genes encode subunits of the pyruvate dehydrogenase enzyme complex, which functions to convert pyruvate to acetyl-coenzyme A (CoA) and carbon dioxide. The fumarate reductase gene frdA also shows significant upregulation. The fumarate reductase complex (frdABCD) is known to function as an anaerobic phosphorylative electron transport chain in other bacteria. Our results suggest that fumarate reductase may also play an important role in the metabolism of M. tuberculosis under starvation conditions.

Lipid biosynthesis

Of the 63 lipid biosynthesis genes on the array, 10 (16%) are significantly downregulated, and only three are up-regulated. The single fatty acid synthase (FAS) I type gene (fas), which generates CoA esters from acetyl-CoA primers, creating precursors for elongation by all other fatty acid and polyketide systems, is downregulated. Several genes involved in the modification of fatty and mycolic acids are also repressed. Both desA3 and desA2, encoding putative aerobic desaturases, are downregulated. cmaA1 and umaA2 are also downregulated and encode proteins involved in the modification of unsaturated meromycolic acid. umaA2 has recently been found to be required for virulence in M. tuberculosis (Glickman et al., 2000).

The genes of the antigen 85 complex, enzymes possessing mycolyl transferase activity and thought to be involved in the transfer of mycolic acids to the cell wall arabinogalactan (Daffe, 2000), are differentially expressed in the starved cultures. Both fbpA and fbpB are downregulated, but fbpC2 is significantly upregulated at the earlier time points only.

Polyketide and non-ribosomal peptide synthesis

All the downregulated genes within the polyketide and non-ribosomal synthesis functional class lie within the genomic cluster including genes encoding the polyketide synthase type I system, ppsABCDE and the type II polyketide synthase, mas, which are involved in the synthesis and transport of phthiocerol dimycocerosate (PDIM) that has been shown to be important for the virulence of M. tuberculosis (Camacho et al., 1999; Cox et al., 1999). Thus, it appears that starvation conditions have caused downregulation of the biosynthesis of this cell wall lipid. Both pks10 and pks2 are upregulated, pks2 being required for sulpholipid biosynthesis (Sirakova et al., 2001) and induced upon macrophage infection (Graham and Clark-Curtiss, 1999).

Translation apparatus

Of the 51 genes on the array involved in ribosomal protein synthesis, 30 (59%) are downregulated, 15 of which lie within a cluster of 24 genes spanning Rv0700–Rv0723. This suggests growth rate-dependent regulation of ribosome number, as seen in other bacteria (Keener and Nomura, 1996). Of the 15 genes involved in protein translation and modification, five are significantly downregulated, including the elongation factor genes tuf and efp. Two aminoacyl tRNA synthases (gltS and trpS) are downregulated, consistent with the coupled synthesis of translation factors, tRNA synthases and ribosome components (Grunberg-Manago, 1996). Downregulation of the translation apparatus is in accordance with findings in Escherichia coli, in which carbon starvation resulted in a 20% drop in protein synthesis during the first hour of starvation (Reeve et al., 1984).

RNA synthesis, RNA modification and DNA transcription

Four of the 12 M. tuberculosis sigma factors (sigB, sigE, sigF and sigD) are upregulated, but expression levels of the principal sigma factor, sigA, did not change significantly. sigB has been shown to be induced during the transition from log phase to stationary phase in M. tuberculosis and also in response to various stress conditions (Hu and Coates, 1999; Manganelli et al., 1999). sigE has been shown to be involved in the mycobacterial stress response (Wu et al., 1997) and is also upregulated upon macrophage infection (Graham and Clark-Curtiss, 1999). sigF is upregulated by stress conditions including antibiotics (Michele et al., 1999), by entry into stationary phase in vitro (DeMaio et al., 1996) and during macrophage infection (Graham and Clark-Curtiss, 1999), and an M. tuberculosis strain lacking sigF has been shown to be less virulent in mice (Chen et al., 2000). The five genes upstream of sigF (Rv3287c–Rv3291c) are also upregulated under starvation conditions (Fig. 3). rsbW, which shares significant homology to the B. subtilis antisigma factor and has been shown to regulate M. tuberculosis sigF negatively (Gomez et al., 1997), is upregulated ninefold and may therefore be functioning to sequester the activity of sigF. Rv3288c and Rv3289c are of unknown function and are upregulated 11- and 22-fold respectively. The lat gene, encoding a protein with homology to lysine-ɛ aminotransferase from Nocardia lactamdurans involved in β-lactam biosynthesis, is the third most highly upregulated gene overall (42-fold induction). Rv3291c, which is homologous to transcriptional regulators of the Lrp/AsnC family, is induced 15-fold.

Figure 3.

Response of the sigF operon to starvation.

A. Profile of the relative expression levels of the sigF operon genes across the time course. The results are the average ratios of the fluorescence intensities for the starved cultures compared with log-phase cultures from three independent hybridizations at each of the 4 h, 24 h and 96 h time points for sigF (open triangles), rsbW (open circles), Rv3288c (open squares), Rv3289c (black circles), lat (black squares) and Rv3291c (black triangles).

B. Schematic representation of the sigF genomic region.

Both rpoA and rpoC, subunits of RNA polymerase, are downregulated (rpoB being significantly downregulated at 4 h only). Rifampicin binds to the β-subunit of RNA polymerase and inhibits transcription at the initiation stage (Jin and Gross, 1988; Tavormina et al., 1996). Decreased levels of RNA polymerase itself suggest that transcriptional activity in the starved cells is reduced and would explain the increased resistance of the starved cultures to rifampicin treatment. Another possibility is that binding of sigma factors upregulated under starvation conditions to the RNA polymerase holoenzyme may confer differ-ential sensitivity towards rifampicin, as shown for E. coli (Wegrzyn et al., 1998). A further possibility is binding of ppGpp to the β-subunit of RNA polymerase and sequestration of its activity, as can occur in E. coli (Reddy et al., 1995). ppGpp accumulates as a result of the stringent response (Cashel et al., 1996). In M. tuberculosis, RelA functions as a ppGpp synthase and has been shown to be important in the survival of M. tuberculosis during nutrient starvation (Primm et al., 2000). Expression levels of relA are induced here, confirming that the stringent response may be important in the survival of M. tuberculosis during nutrient limitation.

Regulatory function

The three most highly induced regulatory genes are Rv2034, Rv1152 and Rv3291c. Rv2034 belongs to the ArsR family of regulators and appears to be co-expressed with Rv2035 (of unknown function) and Rv2036 (weak similarity to a Streptomyces protein involved in lincomycin production). Rv1152 bears homology to the GntR family of repressors. Rv3291c belongs to the Lrp/AsnC family and lies within the genomic region upstream of sigF, as discussed above.

whiB2, the closest M. tuberculosis homologue of the Streptomyces whiB gene (Soliveri et al., 2000), is also induced. This gene has been shown to be essential in M. smegmatis and is thought to be involved in septum formation and cell division in this organism (Gomez and Bishai, 2000). Two genes involved in iron uptake regulation are upregulated. furB encodes a protein with homology to the ferric uptake regulatory protein, Fur, from E. coli and has been found to be upregulated by M. bovis BCG during macrophage infection (Li et al., 2001). The upstream gene, Rv2358, encoding a transcriptional regulator is also upregulated. IdeR is highly homologous to DtxR (diphtheria toxin repressor) from Corynebacterium diphtheriae, which regulates toxin and siderophore production, and there is experimental evidence that the M. tuberculosis IdeR protein plays a role in iron and DNA binding (Schmitt et al., 1995; Rodriguez et al., 1999). Of the two-component regulatory systems, only the kdpE gene is induced. Although its function in M. tuberculosis has not been determined, the kdpDE two-component system in E. coli controls expression of the potassium transport system, which is induced under low K+ conditions (Silver, 1996). It is possible that it also responds to potassium limitation in M. tuberculosis.

The serine/threonine protein kinases pknB and pknD and the putative phosphoprotein phosphatase ppp are downregulated. pknD is organized in an operon with pstS, encoding a phosphate-binding protein. pknB and ppp reside within the same operon as pknA, pbpA and rodA. pbpA was also found to be significantly downregulated. pbpA and rodA control the switch between peptidoglycan elongation and septum formation in other bacteria, and it has therefore been speculated that this cell elongation process may be regulated by a phosphorylation/dephosphorylation cascade (Av-Gay and Everett, 2000). pknB has been shown to be expressed in M. tuberculosis growing exponentially in vitro, in a murine macrophage in vitro infection model and in human alveolar macrophages from a patient with tuberculosis (Av-Gay et al., 1999), circumstances in which the bacterium is likely to be growing actively. It is possible that the expression of these genes is downregulated when M. tuberculosis enters a state of limited metabolism and little or no cell turnover, as seems to occur during starvation.

Cell processes

The majority of differentially expressed cell process genes belong to the transport/binding protein subclass. The four genes subI–csyT–cysW–cysA, highly homologous to the various components of the sulphate transporters of E. coli, are all upregulated, suggesting that these genes may be important to M. tuberculosis under conditions of sulphur limitation. Two of the genes encoding the putative dipeptide transporter dppABCD are induced. The transport of peptides into the cell may be important for the survival of M. tuberculosis during nutrient limitation. The B. subtilis dipeptide transport system is expressed early during sporulation in response to nutrient starvation (Mathiopoulos et al., 1991). However, genes encoding components of the putative phosphate transporters are downregulated. pstB, phoS1 and pstA2 lie within the same operon, whereas pstA1 lies within a second operon in the same genomic region. As the cultures were starved in PBS, phosphate was not limiting and, therefore, these genes would not be expected to be induced.

The parA and parB genes are downregulated. As these genes encode proteins involved in chromosome partitioning (Wheeler and Shapiro, 1997), rates of cell division may be reduced in starved M. tuberculosis. The tig gene is downregulated. We have found that the corresponding protein disappears from gels of 6-week-starved cultures compared with log-phase cultures (Fig. 2; Table 1).

Cell envelope

The most highly upregulated cell envelope gene is pirG. This gene encodes a repetitive protein, which resulted in impaired multiplication within macrophages when disrupted in M. tuberculosis (Berthet et al., 1998). The gene encoding the MPT70 antigen is also highly induced; however, the mpt64 gene is downregulated. The MPT64 protein was also seen to disappear on gels of the 6-week-starved cultures when compared with log-phase cultures (Fig. 2; Table 1).

Unknowns and conserved hypotheticals

Rv2660c is the most highly induced gene overall. Additionally, the following three genes, Rv2661c, Rv2662 and Rv2663, are all upregulated. This group of genes encodes small proteins, all of unknown function, that are present in M. bovis but not in M. leprae. Rv0116c, Rv2558 and Rv1284 are also highly induced. Rv0116c, encoding a 26.9 kDa protein, bears homology to a putative secreted protein from M. leprae. The proteins corresponding to Rv2558 and Rv2557 (also upregulated) were seen to be highly induced on 2D gels of 6-week-starved cultures (Fig. 2; Table 1). These genes have also been found to be expressed within human tuberculous granulomas by in situ hybridization (G. Fenhalls, personal communication) and may therefore be important to M. tuberculosis for in vivo survival. Rv1284 encodes an 18.2 kDa protein that bears homology to a gene of unknown function from Streptomyces coelicolor (TrEMBL accession no. Q9S2W3) and weak similarity to several carbonic anhydrases. Rv1284 lies directly upstream of cysD and cysN, which form the two subunits of a sulphate adenylate transferase, required for the activation of inorganic sulphate. cysD is upregulated under nutrient starvation and has also been shown to be upregulated during macrophage infection (Triccas et al., 1999).


Building upon the early work of Loebel et al. (1933b), we have established an in vitro model, in which nutrient starvation causes M. tuberculosis to arrest growth, minimize aerobic metabolism and become resistant to existing drugs while maintaining viability. Nutrient starvation may therefore mimic some of the features of M. tuberculosis during the persistent state. Owing to its simplicity, reproducibility and ease of handling, the model may be useful for testing novel antimycobacterial drugs aimed at persistent bacteria. Microarray profiling has provided evidence for induction of the stringent response and downregulation of aerobic respiration, translation, cell division and lipid biosynthesis in response to starvation. Induced genes, such as alanine dehydrogenase and fumarate reductase, may facilitate the survival of M. tuberculosis under these conditions and, as such, may represent drug targets of relevance to the eradication of persistent organisms. Many of the induced genes, including Rv2557 and Rv2558, which were also increased at the protein level, are of unknown function. These may perform specific mycobacterial functions and warrant further investigation as to their relevance to latent disease.

Experimental procedures

Bacterial culture and starvation conditions

Cultures were taken from a frozen seed stock of M. tuberculosis H37Rv (NCTC 7416). Seeds were thawed, grown to late log phase without shaking and then diluted 1:100 into a roller bottle in 100 ml of Middlebrook 7H9 media supplemented with 0.2% (v/v) glycerol, 10% (v/v) albumin– dextrose–catalase (ADC) and 0.025% (v/v) Tween 80 at 37°C with constant rolling at 2 r.p.m. After 7 days growth to log phase, cultures were pelleted and washed twice with PBS before being resuspended in PBS, transferred to standing flasks or microtitre plates and incubated at 37°C. For viability determination during starvation, bacteria were cultured in 10 ml volumes in 30 ml bottles (Nalgene), and the number of cfu ml–1 was determined by plating serial dilutions onto 7H10 agar from triplicate cultures at several time points (day 0, weeks 1, 2, 3 and 6). A visual indication of oxygen depletion was gained by the addition of sterile methylene blue solution (500 μg ml–1) to a final concentration of 1.5 μg ml–1 to 10 ml standing cultures maintained under starvation conditions or in Middlebrook 7H9 supplemented as above. Control flasks containing either PBS or Middlebrook 7H9 medium and methylene blue but no bacteria were also set up.

Drug treatment of cultures

Mycobacterium tuberculosis was cultured under starvation conditions in 1 ml aliquots in a 48-well plate for 6 weeks in a humid CO2 incubator. Two millilitres of a 7-day-old log-phase culture of M. tuberculosis was diluted into 48 ml of Middlebrook 7H9 media supplemented as above, and 1 ml was aliquoted per well of a 48-well plate. Rifampicin, isoniazid and metronidazole were added to duplicate wells of the starved and log-phase cultures at final concentrations of 1 and 10 μg ml–1. Control wells for both starved and log-phase cultures received no drug. Cultures were incubated with or without drug at 37°C for 7 days before being harvested by centrifugation, resuspended in 7H9 medium and serial dilutions plated on 7H10 agar to determine bacterial viability.

Two-dimensional gel electrophoresis and protein identification

Mycobacterium tuberculosis cultures starved in PBS for 6 weeks and PBS-washed log-phase cultures were lysed and proteins separated by 2D gel electrophoresis as described previously (Betts et al., 2000). Proteins were focused using both pH 3–10 (non-linear) and pH 4–7 IPG strips (Amersham Pharmacia Biotech), and each sample was run in duplicate. Gels were silver stained (Bjellqvist et al., 1993) and images digitized by scanning on a flat-bed scanner (Epson GT9000). For subsequent identification of protein spots by mass spectrometry, gels were rerun and stained with a silver stain compatible with mass spectrometry (Shevchenko et al., 1996; Betts and Smith, 2001). Protein spots of interest were excised from the gels and tryptic digests prepared for analysis by MALDI MS as described previously (Betts et al., 2000). Proteins were identified by peptide mass fingerprinting using PEPSEA software (Protana). Proteins not identified by MALDI MS were analysed further by LC/MS/MS using a Q-TOF instrument (Micromass) as described previously (Rowley et al., 2000). Spectra were interpreted and proteins identified using MASCOT software (Matrix Science).

RNA isolation

Mycobacterium tuberculosis was cultured under starvation conditions in 30 ml volumes in 150 ml bottles (Nalgene). Three 30 ml cultures were harvested by centrifugation after 4 h, 24 h and 96 h of starvation. Control samples were prepared by washing 7-day-old log-phase cultures twice with PBS then resuspending in PBS, as described for the preparation of starved bacteria, and harvesting 3 × 30 ml by centrifugation at time zero (t = 0). Pellets were resuspended in 1 ml of TRIzol (Life Technologies). The cell suspensions were immediately transferred to tubes containing glass beads (Sigma) and lysed in a ribolyser (Hybaid) for three cycles (20 s at speed 6) with cooling on ice for 1 min between pulses. Samples were then centrifuged (13 000 r.p.m. for 45 s), and the solution above the beads and cellular debris was removed to a tube containing 200 μl of chloroform and Phase Lock gel (heavy; Sigma). Samples were inverted rapidly for 15 s and then periodically for 2 min before centrifugation (13 000 r.p.m. for 5 min). The top aqueous layer was transferred to a fresh tube, and an equal volume of isopropanol was added and mixed well. Precipitated nucleic acids were collected by centrifugation (12 000 g for 20 min at 4°C), and the pellets were washed with 70% ethanol and air dried. Crude RNA samples were treated with DNase I (Ambion) by incubation at 37°C for 20 min and purified further using an RNeasy kit (Qiagen) according to the manufacturer’s instructions. The quality of purified total RNA was assessed by gel electrophoresis.

Preparation of labelled cDNA from total RNA

Fluorescently labelled cDNA copies of total RNA were prepared by direct incorporation of fluorescent nucleotide analogues during a first-strand reverse transcription (RT) reaction. Each 25 μl labelling reaction included 5 μg of RNA, 60 μM random primers, 0.5 mM each dATP, dGTP and dTTP, 0.05 mM dCTP, 10 mM dithiothreitol (DTT) and 200 units of reverse transcriptase (Superscript II; Stratagene) in a 1× reaction buffer provided by the enzyme manufacturer and 2 nmol of either Cy3-dCTP or Cy5-dCTP (Amersham Pharmacia Biotech). The RNA and primers were preheated to 70°C for 10 min and snap cooled on ice before adding the remaining reaction components. The RT reaction was allowed to proceed for 5 min at 25°C followed by 90 min at 42°C. Pooled t = 0 RNA was labelled with Cy3, and triplicate RNA samples taken at later time points were labelled with Cy5. Starved samples taken during the time course were directly compared relative to the pooled t = 0 control by co-hybridization. The two labelled cDNA samples to be compared were combined, ethanol precipitated, dried and then resuspended in RNase-free water.

Microarray hybridization and data analysis

DNA microarrays used consisted of 3649 PCR-amplified ORF-specific DNA fragments, representing 93% of the predicted 3924 M. tuberculosis H37Rv ORFs (Cole et al., 1998). Details of the amplicon generation, array production and quality control testing will be described elsewhere (M. Wilson et al., manuscript in preparation). Combined probes were applied to the array in a hybridization mixture containing 4× SSC, 0.2% SDS, 42% formamide and 0.1 μg μl–1 salmon sperm DNA. Samples were first denatured by heating to 98°C for 3 min, and hybridization was carried out under a glass coverslip in a humidified slide chamber (Corning) submerged in a 42°C water bath for ≈ 16 h. Coverslips were removed in wash buffer I (1× SSC, 0.06% SDS) prewarmed to 37°C, and slides were then washed sequentially in buffer I and buffer II (0.06× SSC) for 2 min each at room temperature before being dried under a stream of nitrogen. Slides were scanned using a ScanArray 3000 instrument (GSI Lumonics), and the resulting images were analysed using GENEPIXPRO 3.0 software (Axon Instruments). Local background was subtracted from the value of each spot on the array. Spots with a value less than local background were excluded from further analyses. The two channels were standardized with respect to the geometric mean per slide for the remaining set of M. tuberculosis DNA spots, and Cy5/Cy3 fluorescence ratios were calculated from the standardized values. Analysis of variance (ANOVA) was used to assess whether the expression ratio at each time point was significantly different from 1, i.e. no change in gene expression relative to the control. The response used in the analysis was the log-transformed ratios calculated from the standardized values. The model fitted in ANOVA was: E(Yij) = αi + βi*timej for each gene i at time point j. The expression ratio at each time point was tested for a difference from 1 using the mean square error (between-slide variation) from the overall model. Resulting P-values range from 1.0 for gene expression levels equivalent in the test and control to very small P-values for expression level differences that are highly significant. Levene’s test was used to test the assumption of equal variance between each of the time points (Levene, 1960). The Shapiro–Wilk statistic was used to test the assumption of normality for each gene (Shapiro and Wilk, 1965).


We gratefully acknowledge Dr Lu Yu for performing the LC/MS/MS analyses, and Dr Mike Wilson for help and advice in setting up the microarray experiments. We also thank Dr Martin Everett for careful reading of the manuscript and helpful advice.

Supplementary material

The following material is available from


Table S1. Mycobacterium tuberculosis H37Rv genes upregulated after 4 h starvation.

Table S2. Mycobacterium tuberculosis H37Rv genes upregulated after 24 h starvation.

Table S3. Mycobacterium tuberculosis H37Rv genes upregulated after 96 h starvation.

Table S4. Mycobacterium tuberculosis H37Rv genes downregulated after 4 h starvation.

Table S5. Mycobacterium tuberculosis H37Rv genes downregulated after 24 h starvation.

Table S6. Mycobacterium tuberculosis H37Rv genes downregulated after 96 h starvation.