The use of microarray analysis to determine the gene expression profiles of Mycobacterium tuberculosis in response to anti-bacterial compounds

Abstract

The response of Mycobacterium tuberculosis to six anti-microbial agents was determined by microarray analysis in an attempt to define mechanisms of innate resistance in M. tuberculosis. The gene expression profiles of M. tuberculosis after treatment at the minimal inhibitory concentration (MIC) for 4 h with isoniazid, isoxyl, tetrahydrolipstatin, SRI#221, SR1#967 and SR1#9190 were compared to untreated M. tuberculosis. A common response to drug exposure was defined, and this expression profile overlapped with a number of other mycobacterial stress responses recently identified by microarray analysis. Compound-specific responses were also distinguished including a number of putative transcriptional regulators and translocation-related genes. These genes may contribute to the intrinsic resistance of M. tuberculosis to anti-microbial compounds. Further investigation into these mechanisms may elucidate novel pathways contributing to mycobacterial drug resistance and influence anti-mycobacterial drug development strategies.

Introduction

In 1993 the World Health Organisation declared tuberculosis (TB) a global emergency.¹ Ten years on, it has been estimated that the global incidence rate of tuberculosis is growing at approximately 0.4%/year.² Of the world's new tuberculosis cases approximately 3% were attributed to multidrug-resistant tuberculosis (MDR-TB) in 2000. Although multidrug-resistant tuberculosis may not be a problem globally, MDR-TB is at critical levels in many hot spots across the world.³ The emergence of MDR-TB, the deadly link between TB and HIV infection, the problems of treatment expense and patient compliance, and the requirement to eliminate persistent infection emphasises the need for new anti-mycobacterial compounds to be developed.⁴

Mechanisms of drug resistance in Mycobacterium tuberculosis have been identified to all five first line anti-mycobacterial drugs—isoniazid (INH), rifampin, pyrazinamide, ethambutol, and streptomycin. M. tuberculosis multiple resistance in these instances is conferred by a series of chromosomal mutations.⁵ However, less is known about mechanisms of intrinsic/natural drug resistance in M. tuberculosis such as reduced cell wall permeability, efflux systems, or the expression of drug-inactivating enzymes. The poor action of many antibiotics, and the relative resistance of bacilli to drying, alkali and many chemical disinfectants has often been attributed to the low permeability of the unusual cell wall structure of mycobacteria.⁶ In addition to the hydrophobic barrier of the mycobacterial cell wall, several genes encoding putative drug efflux systems have been identified in mycobacteria. The probable efflux protein efpA has been reported to be present in slow-growing pathogenic mycobacteria.⁷ The efflux pump LfrA has been demonstrated in M. smegmatis to confer low-level resistance to fluoroquinolones⁸ and to contribute to ethidium bromide resistance,⁹ whereas the M. tuberculosis P55 multidrug efflux pump has been identified to confer aminoglycoside and tetracycline resistance.¹⁰ Indeed, the H37Rv M. tuberculosis genome sequencing project revealed the presence of up to 24 members of the major facilitator superfamily of transporters, and over 80 putative members of the ABC-transporter family.¹¹ Of the ABC-transporter family, 21 export systems were defined in M. tuberculosis, many of which are implicated in the export of drugs and which may contribute to the innate resistance of mycobacteria to broad spectrum antibiotics.¹² M. tuberculosis has also been demonstrated to express β-lactamases¹³ and aminoglycoside acetyltransferases,¹⁴ which may reduce the effectiveness of β-lactams and aminoglycosides against M. tuberculosis. Further understanding of the mechanisms of intrinsic resistance to antibiotic compounds in M. tuberculosis may help to improve existing drug treatments and define new drug development strategies.

The advent of microarray technology has allowed the transcriptional profiles of bacteria to be examined in response to various stresses. The use of M. tuberculosis microarrays was first reported to describe the induction of M. tuberculosis genes in response to INH treatment.¹⁵ Genes were identified encoding proteins related to the mode of action of INH, such as acpM (coding for an acyl carrier protein), kasA and kasB (encoding β-ketoacyl synthases). Other genes most likely involved in the mycobacterial response to the toxicity of the drug were also highlighted—efpA (coding for a putative efflux protein), and aphC (alkyl hydroperoxide reductase, involved in the oxidative stress response). Microarray analysis of gene expression has also recently been used to predict the common functional category of unknown anti-mycobacterial drugs as part of a pipeline of drug discovery.¹⁶

We describe here the use of a gene-specific M. tuberculosis microarray to compare the transcriptional response of M. tuberculosis H37Rv to six compounds with anti-mycobacterial activity: (i) INH, a front line anti-tuberculosis drug targeting mycolic acid synthesis;¹⁷ (ii) Isoxyl, a drug used in the past to treat tuberculosis, targeting a delta-9 oleic acid desaturase and mycolic acid synthesis;^18,19 (iii) Tetrahydrolipstatin (THL), a lipase inhibitor used in the treatment of obesity;²⁰ and three compounds from The Southern Research Institute (Birmingham, Alabama, USA), exhibiting potent anti-mycobacterial properties (iv) SRI#221; (v) SRI#967; and (vi) SRI#9190. The comparison of these six distinct transcriptional profiles defines a common M. tuberculosis response to these anti-mycobacterial compounds, and describes drug-specific changes which may reflect the mode of action of each drug. This investigation also distinguishes genes of unknown function that may contribute to the intrinsic resistance of M. tuberculosis to anti-microbial agents.

Materials and Methods

Growth conditions and RNA extraction

M. tuberculosis strain H37Rv was grown at 37°C in Dubos liquid medium, supplemented with bacto Dubos medium albumin (Becton Dickinson). Mid-log phase mycobacterial cultures were concentrated to 1/20th volume of liquid medium and incubated overnight to recover. Anti-microbial compounds were added at approximately ×1 MIC (determined using the Microplate Alamar Blue Assay, MABA,²¹ at The Southern Research Institute); the structures and MICs of the compounds used are detailed in Fig. 1. The drug-treated, together with untreated control mycobacterial cultures, were incubated at 37°C for 4 h. M. tuberculosis RNA was extracted using the GTC/TRIzol^® method developed by Mangan et al.²² The RNA samples were DNAsel treated and cleaned up on RNeasy^® Mini-Columns (Qiagen).

Figure 1.

The structures of the anti-microbial compounds used in this investigation. The source and the approximate MIC of the compounds against M. tuberculosis are also detailed. MICs were determined using the Microplate Alamar Blue Assay (MABA)²¹ at The Southern Research Institute (Birmingham, Alabama, USA).

Microarray hybridisation and normalisation strategies

cDNA derived from three separate RNA extractions for each of the compounds tested and from untreated control samples were hybridised to a gene-specific PCR product H37Rv M. tuberculosis microarray, the design and generation of which is described in Stewart et al.²³ Details of the M. tuberculosis microarray used in this investigation can be found at http://bugs.sghms.ac.uk/. Two colour competitive hybridisations were performed as previously described Stewart et al.²³ hybridising the mycobacterial RNA-derived cDNA against M. tuberculosis genomic DNA. The hybridised slides were scanned sequentially at 532 and 635 nm corresponding to Cy3 and Cy5 excitation maxima using the 4.28™ Array Scanner (Affymetrix). Comparative spot intensities from the images were calculated using Imagene 4.0 (BioDiscovery), and imported into GeneSpring 4.2 (Silicon Genetics) for further analysis. The array data was normalised to the 50th percentile, and values of less than zero were adjusted to zero. Repeat hybridisations using the same cDNA samples (between 3 and 7 replicates for each condition) were replicated together. The experiments were then normalised to the untreated control sample using a per gene normalisation strategy.

Microarray data analysis

Two measures of significance were applied to the normalised data set to identify differentially regulated genes (i) a minimum p-value of 0.05 incorporating the cross-gene error model (GeneSpring) was set to discriminate genes significantly deviating from the 1:1 ratio (treated : untreated) which were then subjected to Benjamini and Hochberg correction to take into account multiple experiment testing and (ii) a one-way ANOVA (GeneSpring).

A technique of single spot replacement, SSR (J. Bacon, personal communication) was also used to enhance the original data set. The un-normalised cDNA : genomic DNA ratios for each replicate under each condition were imported into Microsoft Excel. For each element on the microarray the individual ratio furthest from the median of the replicates was replaced with the mean of the remaining ratios. In this way the effect of extreme values was minimised from the data set. This SSR data set was then normalised as previously described, and subjected to two measures of significance: (i) the statistical group comparison (ANOVA); and (ii) the statistical package SAM (Significance Analysis of Microarrays, version 1.15²⁴) was used to identify genes differently expressed in the normalised data sets. A minimum fold change of 1.5 between control and drug-treated data sets, and a false discovery rate (FDR) of less than one (of the median) was used as a measure of significance.

The hypergeometric distribution was used to determine if particular functional categories of genes were enriched in response to each drug treatment. The hypergeometric p-values were calculated as described by Boldrick et al.²⁵ where N=3924 the total number of genes in the population, A= the number of genes within each functional classification, x= the number of genes identified as up-regulated in response to each drug, and n= the total number of genes up-regulated after treatment by each anti-microbial compound.

Results

The transcriptional response of M. tuberculosis to each of the six anti-microbial agents was defined as the subset of genes identified as significantly differentially expressed in two or more statistical tests (described in the Methods). Using this analysis strategy, 155 genes were demonstrated to be up-regulated by INH treatment (32 down-regulated); Isoxyl (231 up, 21 down); Tetrahydrolipstatin (208 up, 24 down); SRI#221 (182 up, 25 down); SRI#967 (116 up, 30 down); and SRI#9190 (124 up, 22 down). The fold changes and predicted function of these genes are described as supplementary information Table S1.

Dissecting the transcriptional response of M. tuberculosis to the six drug compounds by functional classification (as described by Cole et al.¹¹) revealed that the genes induced by drug treatment broadly represent most of the range of pathways that are present in M. tuberculosis. The hypergeometric distribution²⁵ was used to determine whether the enrichment of genes within a particular functional category in response to each drug treatment was significant (p-value <0.05). Table 1 shows that the number of genes within the functional categories of energy metabolism and chaperones/heat shock were significantly enhanced after treatment with each of 3 or more anti-microbial agents. The functional category of lipid metabolism was significantly enriched in response to INH and isoxyl treatment, as was the category of polyketide synthesis after treatment with SRI#967. Additionally, the proportion of genes involved in the metabolism of the cell envelope was significantly increased after treatment with SRI#221 and SRI#967 (Table 1).

Table 1.

The M. tuberculosis response to 6 anti-microbial compounds examined by functional category, as defined by Cole et al.¹¹ Only genes identified as significantly over-expressed in two or more of the statistical tests described in response to isoniazid (INH), isoxyl (ISO), tetrahydrolipstatin (THL), SRI#221/967/9190 are detailed in these tables. These gene lists are described in Supplementary Table S1. The hypergeometric probabilities²⁵ of the enrichment of particular functional categories of genes in response to each drug treatment are indicated if ^ap-value <0.05, ^b<0.01, ^c<0.001.

Functional classification (as defined by Cole et al.11)	Genes identified as up-regulated
	INH	ISO	THL	221	967	9190
I. Small-molecule metabolism
A. Degradation (163)	8	8	8	5	8	7
B. Energy metabolism (292)	10	32c	18	18a	9	14a
C. Central intermediary metabolism (45)	3	3	1	4		1
D. Amino acid biosynthesis (95)	4	7	1	2	2	2
E. Polyamine synthesis (1)
F. Purines, pyrimidines, nucleosides and nucleotides (60)	2	2		5
G. Biosynthesis of cofactors, prosthetic groups and carriers (117)	5	2	4	7	2	3
H. Lipid biosynthesis (65)	7b	11c	4	5	3	3
I. Polyketide and non-ribosomal peptide synthesis (41)	4	3	2	3	4a	2
J. Broad regulatory functions (187)	4	5	7	9	8	5
II. Macromolecule metabolism
A. Synthesis and modification of macromolecules (215)	10	17a	17a	12	8	10
B. Degradation of macromolecules (87)	3	7	5	5	1	2
C. Cell envelope (360)	16	23	18	26b	20b	9
III. Cell processes
A. Transport/binding proteins (123)	7	7	9	4	1	3
B. Chaperones/heat shock (16)	3a	6c	3a	1		3a
C. Cell division (19)		2	3	1
D. Protein and peptide secretion (14)		2	1	1		1
E. Adaptations and atypical conditions (12)					1
F. Detoxification (22)		1	1	3	1	1
IV. Other
A. Virulence (38)			1	1	1
B. IS Elements, repeated sequences and phage (135)	9a	1	11a	1	3
C. PE and PPE families (167)	10	6	9	2		4
D. Antibiotic production and resistance (14)			1	1		1
E. Bacteriocin-like proteins (3)				1
F. Cytochrome P450 enzymes (22)	1	1	2	3	1	1
G. Coenzyme F420-dependent enzymes (3)		1
H. Miscellaneous transferases (61)	1	3	4	2	3	2
I. Miscellaneous phosphatases, lyases and hydrolases (18)						1
J. Cyclases (6)			1	1
K. Chelatases (2)
V. Conserved hypotheticals (915)	36	61	40	26	29	30
VI. Unknowns (606)	12	20	37	33	11	19
Total genes (3924)
Total up-regulated genes after drug treatment	155	231	208	182	116	124
Percentage of unknown function (39% of total genes unknown)	31	35	37	32	34	40

Common response to anti-mycobacterial agents

By comparing the similarities between the M. tuberculosis drug-induced expression profiles, a common response to anti-mycobacterial agents could be defined. A subset of 80 genes were identified which were significantly up-regulated after treatment with 3 or more anti-microbial compounds (of a maximum 6). These genes are listed in Table 2. Many of these common genes induced by exposure to anti-microbial compounds were involved in the mycobacterial stress response. Genes associated with DNA repair such as end (coding for a probable endonuclease) and recA (encoding recombinase A²⁶) were up-regulated; together with Rv3049c (a probable monoxygenase) and aphC (alkyl hydroperoxide reductase) expressed in response to oxidative stress.²⁷ Also over-expressed after drug treatment (>3 drugs) were gltA1 (a probable citrate synthase) and icl (isocitrate lyase) similar to changes in metabolism seen under stress conditions.²⁸ RNA polymerase sigma factors A and B were also induced, together with serine/threonine protein kinases B and G. sigB has been implicated in the M. tuberculosis response to a number of stress conditions.²⁹ The product of pknG is predicted to be a soluble protein (the transcription of which may be controlled by the redox status of the cell) which may be involved in glutamine uptake and which may be up-regulated under nitrogen-limiting conditions.³⁰ Additionally the putative nitrate/nitrite transporter narK2, the nitroreductase acg and the nitrate reductase narH were also identified to be induced. Indeed, five genes demonstrated to be part of the ACG (acr-coregulated gene) family were found to be up-regulated after exposure to anti-mycobacterial compounds—Rv1733c, narK2, Rv1738, Rv2005c, and acg.³¹

Table 2.

The common response of M. tuberculosis to 6 anti-microbial agents, detailing the genes identified to be up-regulated in response to 3 or more of the anti-microbial compounds tested (maximum 6).

N	Gene name	Rv No	INH	ISO	THL	221	967	9190	Putative function	A	B	C	D	E	F	G	H	I	J
3	pknB	Rv0014c				2	3	3	Serine/threonine protein kinase B								•
4	rpsF	Rv0053		2		2	3	3	Ribosomal protein S6			•
5	icd2	Rv0066c		2	2	3	3	3	Isocitrate dehydrogenase			•					•
3	fbpC2	Rv0129c		4	2			3	Antigen 85c, mycolyl transferase C	•			•
4	pntAB	Rv0156	2	3	2	3			Probable NAD(P) transhydrogenase
4	bglS	Rv0186	2	2			2	3	Probable beta-glucosidase								•
4	Rv0247c	Rv0247c		2	3		3	2	Probable succinate dehydrogenase
3	fadD2	Rv0270	4		2			2	Probable long-chain fatty acid CoA ligase
3	Rv0349	Rv0349	2		2	3			Unknown
3	pknG	Rv0410c		2	2	2			Serine/threoine protein kinase G			•
3	Rv0412c	Rv0412c	2	2		2			Possible conserved membrane protein			•				•
3	Rv0446c	Rv0446c		3		2	2		Possible conserved membrane protein						•
3	icl	Rv0467	4	2				2	Isocitrate lyase						•	•	•		•
3	mmpS2	Rv0506				2	2	3	Unknown, probable membrane protein
3	end	Rv0670	2	2				2	Probable endonuclease IV			•	•			•
4	rpsS	Rv0705		2	2		3	4	30s ribosomal protein s19
4	rpmC	Rv0709		3	2		2	4	50s ribosomal protein L29
3	rplR	Rv0720	4	2				2	50s ribosomal protein L18
3	Rv0851c	Rv0851c				2	2	2	Probable dehydrogenase/reductase						•
4	fadE10	Rv0873	4	4	2		2		Probable acyl-CoA dehydrogenase							•	•
3	Rv0910	Rv0910		2	2	2			Unknown
3	sucC	Rv0951	3	3	2				Probable succinyl-CoA synthetase		•
4	esxl	Rv1037c	4	3			3	3	Putative ESAT-6-like protein								•
5	Rv1109c	Rv1109c	2	3		3	3	2	Unknown							•
3	gltA1	Rv1131	4	2	2				Probable citrate synthase						•	•	•		•
4	narH	Rv1162	2	2		2		2	Probable respiratory nitrate reductase			•					•
3	papA3	Rv1182			3		3	3	Polyketide associated protein				•					•
3	Rv1184c	Rv1184c		2	3		3		Unknown, possible exported protein						•
4	esxK	Rv1197	3	2			3	3	Putative ESAT-6 like protein
4	esxL	Rv1198	4	4			2	4	Putative ESAT-6 like protein
3	atpG	Rv1309		2	2	2			ATP synthase gamma chain
3	PPE19	Rv1361c	4	4			2		PPE family protein
4	appC	Rv1623c	2	2		3		3	Probable cytochrome D ubiquinol oxidase
4	Rv1683	Rv1683	2		2		2	2	Possible long-chain acyl-CoA synthase			•
4	Rv1733c	Rv1733c	4	4			3	4	Probable conserved transmembrane protein			•		•	•
3	narK2	Rv1737c	4	3	2				Possible nitrate/nitrite transporter			•		•	•
4	Rv1738	Rv1738		4		2	2	3	Unknown			•		•	•			•
3	Rv1747	Rv1747		2	2		2		Probable membrane transport protein
4	esxN	Rv1793	3	3			3	3	Putative ESAT-6-like protein
4	Rv1813c	Rv1813c	2	4	2			2	Unknown			•		•	•
4	Rv1987	Rv1987	3		4	3	3		Probable chitinase		•		•
3	ctpF	Rv1997		2	2			2	Probable metal cation transporter			•		•	•
4	Rv1998c	Rv1998c		2	2		2	2	Unknown			•		•
3	Rv2005c	Rv2005c	4	4	2				Unknown			•		•	•			•
3	fdxA	Rv2007c	2	4				2	Probable ferredoxin		•	•		•	•			•	•
6	acg	Rv2032	3	4	2	3	3	4	Unknown, possible nitroreductase			•		•	•
4	Rv2091c	Rv2091c		3		2	3	4	Unknown, probable membrane protein				•			•
3	Rv2147c	Rv2147c		3	2			3	Unknown protein
3	Rv2185c	Rv2185c		2		2		2	Unknown (TB16.3)				•					•
5	cbhK	Rv2202c	3	3	2		2	3	Probable carbohydrate kinase								•
4	fabD	Rv2243	2	4	2	3			Malonyl CoA-acyl carrier transacylase	•	•	•
5	acpM	Rv2244	4	4		4	3	3	Meromycolate extension acyl carrier protein	•	•	•
4	kasA	Rv2245	2	2		2	2		Beta-ketoacyl-ACP synthase	•	•		•
3	htpG	Rv2299c		4	3	3			Probable heat shock protein									•
4	esxO	Rv2346c	3	3			2	4	Putative ESAT-6 like protein
3	Rv2405	Rv2405	2	2		3			Unknown				•
3	ahpC	Rv2428			3	4	3		Alkyl hydroperoxide reductase C	•	•	•			•				•
4	pepD	Rv2467	3	4	2		3		Probable aminopeptidase (pepN)
5	Rv2626c	Rv2626c	3		2	2	2	2	Unknown			•		•	•
3	Rv2627c	Rv2627c		4		2	2		Unknown			•		•	•
3	Rv2629	Rv2629			2	2		2	Unknown			•		•	•
3	sigA	Rv2703				3	2	3	RNA polymerase sigma factor A								•
4	sigB	Rv2710	3		2	2		4	RNA polymerase sigma factor B				•		•	•	•	•
3	recA	Rv2737c		2	2			3	Recombinase A protein
6	35kd_ag	Rv2744c	2	2	2	2	2	3	35-kd alanine rich antigen						•	•	•	•
4	thyX	Rv2754c		2		2	2	3	Probable thymidylate synthase
3	Rv2818c	Rv2818c			3	4	3		Unknown
3	efpA	Rv2846c	4	4	2				Putative efflux protein	•	•
3	Rv2959c	Rv2959c		2	2	3			Possible methyltransferase
5	Rv3049c	Rv3049c		2	2	2	2	2	Probable monoxygenase
3	PPE51	Rv3136	4	2				2	PPE family protein			•
3	rubB	Rv3250c	3	3				2	Probable rubredoxin						•
3	ctpC	Rv3270		2	2			2	Probable metal cation transporter			•				•
3	PE_PGRS52	Rv3388	2	2			2		PE_PGRS family protein
3	PPE60	Rv3478	4	4	2				PPE family protein
3	Rv3592	Rv3592	3	3			2		Unknown (TB11.2)							•
4	panD	Rv3601c	2			2	2	4	Aspartate1-decarboxylase
3	mmpL8	Rv3823c				2	2	2	Conserved large membrane protein
3	papA1	Rv3824c				2	2	2	Polyketide associated protein				•				•
5	esxB	Rv3874		2	2	2	2	2	10 KDa culture filtrate antigen (cfp10)

Numbers in the columns (INH isoniazid, ISO isoxyl, THL tetrahydrolipstatin, SRI# 221/967/9190) indicate the number of statistical tests in which each gene was found to be significantly induced (minimum of 2, maximum 4). The left column labelled N, details the number of drugs in which each gene was significantly differentially expressed. Dots present in the columns A–J indicate that the gene has been previously identified to be up-regulated in response to various other stresses; A INH treatment,¹⁵ B INH and TLM treatment,¹⁶ C low oxygen,³² D nutrient starvation,³⁷ E nitric oxide treatment,³⁵ F phagocytosis,³⁶ G carbon starvation (Hampshire et al., this issue), H detergent stress,³⁴ I heat shock,²³ J acid shock.³³ This table is ordered by Rv number.

Many of the ‘common’ genes induced by drug treatment have been identified as part of the mycobacterial response to other stresses such as low oxygen,³² heat shock,²³ acid shock,³³ detergent stress,³⁴ nitric oxide treatment,³⁵ phagocytosis³⁶ or nutrient/carbon starvation³⁷ Hampshire et al., this issue. Those genes, which have also been previously identified to be up-regulated in response to these other stresses are marked in Table 2. Of the remaining genes induced by 3 or more drugs, five are annotated as efflux proteins or transporters—narK2 (a possible nitrate/nitrite transporter), Rv1747 (a probable ABC transporter), ctpF (a putative metal cation transporter), efpA (an efflux protein), and ctpC (a probable metal cation transporter). A second subset of six genes belonging to the ESAT-6 family of proteins was also identified as up-regulated after 4 h drug exposure (Rv1037c, Rv1197-Rv1198, Rv1793, Rv2346c, and Rv3874). The ESAT-6 gene clusters in M. tuberculosis have been associated with the generation and transportation of T-cell antigens lacking detectable secretion signals.³⁸ These genes linked to the transportation of unknown moieties may be directly involved in the mycobacterial response to drug compounds which contributes to the intrinsic resistance of mycobacteria to anti-microbial agents.

Drug-specific expression responses

The expression profiles of M. tuberculosis treated with each of the six anti-microbial compounds were compared, by generating a similarity matrix detailing the number of overlapping genes between two drug treatments as a proportion of the possible maximum (Fig. 2). The mycobacterial response to INH and isoxyl exposure was most similar as may be expected as both drugs target aspects of fatty acid and mycolic acid biosynthesis.^17,18,19 The expression profiles of M. tuberculosis treated with the compounds SRI#967 and SRI#9190 were also similar. Aspects of the mycobacterial response to individual drugs focusing on the possible action of the compounds is briefly presented below.

Figure 2.

A similarity matrix comparing the genes up-regulated in response to the 6 anti-microbial compounds tested (isoniazid, INH; isoxyl, ISO; tetrahydrolipstatin, THL; SRI#221/967/9190). Numbers in the top half of the figure represent the number of common genes up-regulated in response to two compounds. The maximum number of genes identified as significantly up-regulated (in more than 2 statistical tests) after treatment with each drug are displayed in the shaded cells. The bottom half of the matrix describes the number of genes common to two drug responses as a proportion of the possible maximum, calculated as 1-(common genes/maximum possible genes). The smaller this proportion is the greater the extent of the overlap between expression responses.

INH and isoxyl

Both INH and isoxyl inhibit fatty acid and mycolic acid biosynthesis in M. tuberculosis.^17,18,19 INH has been demonstrated to target the enoyl-AcpM reductase InhA, a component of the fatty acid synthase—II (FAS-II).³⁹ Isoxyl treatment inhibits mycolic acid and shorter-chain fatty acid synthesis leading to the hypothesis that isoxyl may act on other components of FAS-II.¹⁸ Interestingly, it has also recently been shown to target a delta-9 desaturase in mycobacteria.¹⁹ Genes coding for enzymes involved in FAS-II were up-regulated after exposure to both drugs: fabD (coding for a malonyl-CoA::acyl carrier protein (ACP) transferase), acpM (an acyl carrier protein), kasA and kasB (both β-ketoacyl ACP synthases). These have been previously identified to be induced by INH, ethionamide and thiolactomycin treatment.^15,16 Interestingly amongst other fatty acid biosynthetic genes induced by isoxyl treatment alone was mabA, a gene coding for a β-ketoacyl ACP reductase, which also belongs to the FAS-II system. The induction of mabA (which is transcriptionally linked to inhA in M. tuberculosis), after exposure to isoxyl, but not INH, may reflect differences in the mode of action of these two compounds. Further experiments such as the overexpression of mabA during isoxyl exposure may help to elucidate the primary target of isoxyl.⁴⁰

Tetrahydrolipstatin

Tetrahydrolipstatin (THL) is a reversible inhibitor of lipases used in the treatment of obesity (Xenical^®, Roche). A number of M. tuberculosis putative lipases were up-regulated in response to THL treatment—Rv1683, lipD and lipV (although this was not significant by hypergeometric testing). Of the remaining induced genes, 4 encoding putative transporters (ctpl, sugA, Rv3253c, and Rv3781) and 6 coding for probable transcriptional regulators were identified (Rv0043c, Rv0823c, sigE, Rv3167c, Rv3687c and Rv3855). Also up-regulated on exposure to THL were 4 genes located in a gene cluster Rv0676c-Rv0679c. Rv0676c (mmpL5) belongs to a family of conserved large membrane proteins, Rv0677c (mmpS5) is part of a related small membrane protein family (which appears to overlap stop and start codons with mmpL5), the function of Rv0678 is unknown, and Rv0679c codes for a threonine-rich protein of undetermined function. The functional significance of this cluster of genes in the M. tuberculosis response to THL treatment cannot be elucidated by microarray analysis alone. However, further experimentation into this cluster or other genes of interest may define novel mechanisms of drug resistance in M. tuberculosis.

SRI#221

Figure 1 shows that SRI#221 is a potent anti-tubercular compound with a low MIC value, however the primary mode of action of this anti-microbial compound is unknown. Treatment of M. tuberculosis with SRI#221 induced two clusters of genes involved in complex lipid biosynthesis. The first cluster containing tesA (a probable thioesterase), drrB (an ABC-type transporter), papA5 (a polyketide synthase associated protein), and fadD28 (a fatty acid-CoA ligase), is involved in the biosynthesis and translocation of the multi-methyl branched mycocerosic acids in the generation of phthiocerol dimycocerosates.⁴¹ The second gene cluster includes Rv3822 (of unknown function), mmpL8 (a conserved large membrane protein), papA1 (a polyketide synthase associated protein) and fadD23 (a probable fatty acid-CoA ligase). These genes cluster around the polyketide synthase pks2, responsible for the biosynthesis of the multi-methyl branched phthioceranic acids present in the sulpholipid complex lipids.⁴² The induction of these gene clusters may be part of a compensatory network to minimise the anti-mycobacterial effects of SRI#221, may reflect the broad nature of SRI#221 action, or highlight the primary mode of SRI#221 action to be within shared basic lipid biosynthetic pathways.

SRI#967 and SRI#9190

The mode of action of the anti-mycobacterial compounds SRI#967 and SRI#9190 are yet to be determined. The M. tuberculosis response to SRI#967 exposure includes the up-regulation of Rv0076c and Rv0077c. Rv0076c encodes a probable membrane protein, whereas Rv0077c codes for a possible oxidoreductase. Similarly, Rv0135c (a putative transcriptional regulator) and Rv0136 (belonging to the cytochrome P450 group of monoxygenases) were induced by SRI#9190 treatment alone. The induction of these two distinct clusters (on exposure to different anti-mycobacterial compounds) may be part of a similar oxidative stress response. The identification that these genes may play a role in the intrinsic resistance of M. tuberculosis to anti-microbial agents using microarray analysis enables more specific experimental strategies to be employed.

Of particular interest was a cluster of 4 genes which were significantly induced on exposure to both SRI#967 and SRI#9190. Rv3159c (encoding PPE53, a member of the PPE family), Rv3160c (a putative transcriptional regulator), Rv3161c (a probable dioxygenase) and Rv3162c (a possible integral membrane protein) were up-regulated on exposure to these compounds alone (none of these genes were induced by the other anti-microbial agents tested). The probable dioxygenase Rv3161c is most similar to ring hydroxylating dioxygenases, so it is likely that this enzyme is involved in the degradation of benzene ring structures (which SRI#967 and SRI#9190 both contain, Fig. 1). Rv3160c and Rv3161c have recently been identified to be induced by triclosan treatment.¹⁶ Triclosan (2,4,4′-trichloro-2-hydroxydiphenyl ether) contains two chlorinated benzene rings. This cluster of genes may therefore be induced on exposure to compounds containing halogenated benzene rings, this would explain the up-regulation of the cluster in response to SRI#967 and SRI#9190, but not to the benzene ring structures in isoxyl and SRI#221. This gene cluster may be induced as part of a response to render halogenated benzene compounds benign, so contributing to the natural resistance of M. tuberculosis to a range of anti-microbial agents.

Discussion

The response of M. tuberculosis to six anti-microbial compounds was determined by microarray analysis. The microarray expression data set was analysed using several statistical methods, the use of multiple statistical methods added extra depth to the interpretation of the data sets. Additionally, a single spot replacement strategy was used alongside the original data set to allow the maximum amount of information to be extracted from the data sets without significantly shifting the expression patterns.

Using this microarray analysis strategy, elements of a M. tuberculosis common stress response and specific drug-induced changes were identified to six anti-microbial compounds. The up-regulation of genes specific to each compound may reflect the mode of action of the drug or define innate resistance mechanisms in M. tuberculosis. This investigation has defined an initial subset of genes which may be important in the innate resistance of M. tuberculosis to anti-microbial agents. Microarray profiling of M. tuberculosis gene expression after exposure to drugs enables compounds of unknown mechanism of action to be classified into similarity groups, but in this study has not been helpful to elucidate the site of or mechanism of action. This would require complementary genetic and biochemical studies informed by microarray profiling.

Supplementary Data Table S1 (on the web)

The differentially regulated genes in response to 6 anti-microbial compounds. These tables describe the genes identified as significantly differentially expressed in the M. tuberculosis responses to each of the six anti-microbial compounds tested. Each table consists of the gene name (and unique gene identifier), Rv number, a brief description of the proposed function of the gene and the fold expression ratios determined using each of the four statistical tests described. The number of times each gene has been identified as significantly differentially expressed is detailed in the column labelled N. Cells in this column coloured purple indicate the presence of consecutive Rv numbers. Only genes identified by two or more statistical tests have been included in these tables. These tables are ordered by Rv number.