1 Department of Medicine, Division of Infectious Disease and Geographic Medicine, Stanford University School of Medicine, Stanford, CA 94305-5107, USA
2 Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305-5307, USA
3 Statistics Department, Sequoia Hall, Stanford, CA 94305, USA
Correspondence
Peter M. Small
peters{at}gatesfoundation.org
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
*These authors contributed equally to the work and the order of their names was determined by chance.
Present address: Shanghai Medical College, Fudan University, Shanghai 200032, P. R. China.
Present address: Keck Bioinformatics User Center, UCLA, Los Angeles, CA 90095, USA.
Genes listed in Tuberculist that do not appear in our dataset, either because they are not on the microarray, or because they were removed from the dataset due to not enough good data are available in Supplementary Table S1; all genes in the data set, with expression level and functional categories in Supplementary Table S2; identity graphs, in which the log ratios from each replicate array in a set were plotted against the corresponding values from another array of the same strain, as Supplementary Figure S1; a hierarchical cluster diagram of genes as Supplementary Figure S2; a histogram of non-specific hybridization as Supplementary Figure S3 at http://mic.sgmjournals.org.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The genetic diversity of any pathogen requires special consideration when developing drugs, vaccines and diagnostics. There is always the possibility that a reagent may be effective with some strains of an organism and not with others. Therefore it is useful to survey genetic diversity among clinical isolates, in order to obtain some sense of the extent and types of variability that occur. This will help researchers to identify targets that are less likely to vary among strains, perhaps because their functions are essential to the life cycle of the organism. Some insights into M. tuberculosis genetic variation have come from comparison of the two sequenced genomes; the frequency of single nucleotide polymorphisms in M. tuberculosis is lower than in other bacteria (Fleischmann et al., 2002). Because horizontal gene transfer has not been demonstrated in M. tuberculosis, most of the genetic diversity of M. tuberculosis occurs in the form of insertions and deletions, as well as polymorphisms in the PE and PPE genes.
While several groups, including our own, have examined variability of M. tuberculosis at the DNA level, this is the first systematic survey of variability in mRNA expression among clinical isolates of M. tuberculosis. Genes whose expression varies among isolates when assayed under a single growth condition may make poor drug targets and vaccine antigens and may affect molecular diagnostics, so they can be used to narrow down lists of candidate molecules. Because the measurement of gene expression is extremely sensitive to environmental conditions, comparison of gene expression is labour intensive. In this study, we surveyed 12 strains. These strains are a subset of those for which we have already published genomic deletion information (Kato-Maeda et al., 2001). In order to ensure maximum reproducibility of the experiments and avoid complications caused by differences in growth conditions, we measured gene expression under well-controlled in vitro conditions. Our aims were to provide an overview of gene expression variability among clinical isolates under a single growth condition and to test whether gene functional classes are related to variability in expression.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Preparation of M. tuberculosis cell culture.
Frozen aliquots of the isolates were inoculated into 500 ml roller bottles, each containing 30 ml of 7H9 broth supplemented with 0·05 % Tween 80. The roller bottles were rolled at 6 r.p.m. at 37 °C, ambient atmosphere. All cultures were measured daily on the spectrophotometer at 600 µm. The OD600 values of the cultures were used to monitor bacterial growth. To obtain homogeneous cultures (without clumping), the cultures were maintained in exponential phase (OD600 between 0·1 and 0·5) through repeated dilution into fresh media. The subcultures were examined daily to monitor clumping using an inverted microscope at 100x magnification. Once a homogeneous culture was obtained, the culture was diluted to an OD600 of approximately 0·1 and grown for an additional 24 h; then the OD600 was measured again to ensure they were in exponential phase. The culture was diluted to an OD600 of 0·14 and further incubated for an additional 3 h. Finally, the cells were pelleted by centrifugation and quickly frozen on dry ice.
Experimental reproducibility.
For comparison of gene expression, it is essential that RNA be extracted from bacteria under precisely controlled standardized growth conditions, because gene expression patterns are quite sensitive to small variations in growth conditions and phases. In order to compare the gene expression profiles of different clinical isolates, we ensured that the cultures were all grown under the same conditions and collected at the same growth phase. The different bacterial strains entered exponential phase at different times, some 2 days, some 7 days after inoculation. Four separate cultures were prepared from each strain in order to provide replicates reflecting the variability inherent in handling the cells. To evaluate the reproducibility of experiments among the four independent cultures, we performed linear regression using the normalized log2(Ch2 RNA/Ch1 DNA) ratios of one array from each condition as the predictor variable and the three other replicate arrays as the response variable. The R2 values ranged from 0·62 to 0·92. Visual inspection of identity graphs, in which the log ratios from each replicate array in a set were plotted against the corresponding values from another array of the same strain, indicated a clear unskewed linear relationship between the sets of values from the replicate arrays (data available online as Supplementary Figure S1 at http://mic.sgmjournals.org).
Isolation of M. tuberculosis total RNA.
M. tuberculosis total RNA was isolated from 2530 ml culture (OD6000·2) using an Ambion Total RNA Isolation Kit (Ambion Inc.). The manufacturer's protocol was followed, except that cell walls were disrupted using a Mini bead beater (BioSpec Products). After frozen bacterial pellets were resuspended in 1 ml denaturation solution, they were transferred to a 2 ml screw-cap tube containing 0·5 ml glass beads (diameter 0·1 mm). This mixture was shaken in the Mini bead beater six times for 30 s at maximum speed (
5000 r.p.m.) before performing the rest of the extraction.
DNA was removed from 1020 µg of total RNA using the Ambion DNA Free Kit (Ambion Inc.) according to the manufacturer's instructions. The total RNA was loaded on a 1·2 % agarose gel for electrophoresis. Clear bands of 16S and 23S bacterial rRNA subunits indicated little degradation of RNA. The RNA concentration was quantified by spectrophotometry.
Preparation of fluorescent-labelled cDNA and genomic DNA probes.
Fluorescent-labelled cDNA was synthesized by reverse transcription from total RNA in the presence of Cy5-dUTP. The RT reaction was carried out using 2 µg total RNA in 50 mM Tris/HCl (pH 8·3), 75 mM KCl, 5 mM MgCl2, 0·2 mM dNTPs (except 0·02 mM dTTP), 1·5 µl Cy5-dUTP (Amersham), 2 µg N6/N10 random primer and 200 U of Superscript II reverse transcriptase (Gibco-BRL). The RT reaction was incubated for 10 min at 25 °C, followed by 2 h at 42 °C.
H37Rv genomic DNA was digested with AluI, and then labelled by primer extension in the presence of Cy3-dUTP using DNA polymerase Klenow fragment. In the reaction, digested DNA (2 µg) was mixed with N6/N10 random primer (2 µg) in a total volume of 50 µl. The reaction mixture was heated to 95 °C for 5 min, and then cooled on ice. Five microlitres 10x EcoPol buffer (New England Biolabs), 5 µl dNTP (2 mM each dATP, dCTP, dGTP and 0·2 mM dTTP), 1·5 µl Cy3-dUTP (Amersham) and 4 µl DNA polymerase Klenow fragment (5 U/µl, New England Biolabs) were added to the sample. The reaction was incubated at 37 °C for 2 h.
The Cy5-labelled RT reaction products were combined with 1 µl Cy3-labelled DNA sample. The mixture was diluted to 400 µl and then concentrated to 7 µl through a Microcon-YM10 filter (Millipore).
Microarray hybridization.
To the 7 µl combined labelled cDNA and genomic DNA probe, 1·9 µl 20x SSC, 1·52 µl 2 % SDS and 0·67 µl 10 µg yeast tRNA µl1 were added to a total volume of 11 µl. The probes were denatured by heating at 95 °C for 2 min, then placed on the array under a 20 mmx20 mm glass coverslip. The hybridization was carried out overnight at 65 °C in a humidified slide chamber. Arrays were washed by submersion and agitation once in 1x SSC with 0·05 % SDS for 2 min followed by twice in 0·06x SSC for 1 min. The arrays were dried by centrifugation. Glass microarrays spotted with PCR amplicons based on the H37Rv genome sequence were used, as previously described by Behr et al. (1999).
Array quantification and data analysis.
The hybridized array was scanned using an Axon scanner. Cy3 and Cy5 hybridization intensities for each spot were measured using ScanAlyze software (http://rana.lbl.gov/EisenSoftware.htm). Data were submitted to the Stanford Microarray Database (SMD). Raw data are available from SMD at http://genome-www.stanford.edu/microarray. Prior to data analysis, spots whose regression correlation was below 0·5 were removed from analysis (see ScanAlyze documentation for explanation of this quality control parameter), and then spots missing more than 20 % of their data across all the arrays in the set were removed from analysis. Of 3778 unique sequences represented on the array, 3595 were retained after filtering in this way. A list of genes for which good-quality data were not available in this set of experiments is shown online in the Supplementary materials, Table S1 (http://mic.sgmjournals.org).
log2[(Ch2 IntensityCh2 Background)/(Ch1 IntensityCh1 Background)] ratios were used to represent RNA expression levels for each strain. This differs from commonly used experimental designs in which two RNA samples are compared on a single array, and the log ratio represents the fold difference in expression of a given gene between the two samples. When genomic DNA is used in the reference channel, each gene sequence in the DNA sample is present in a molar ratio of 1 compared with all other sequences in this sample. Therefore, dividing the RNA intensity value by the DNA intensity value is effectively the same as dividing it by 1, except that the signal for each gene becomes normalized for spurious printing and hybridization effects. Therefore the log ratio represents the log expression level of the RNA in the sample and not the relative expression levels between two samples.
The scale of the values of the log ratios is arbitrary, as the detection sensitivity for the RNA and DNA samples are set independently and are unrelated. Therefore the assumption was used that the overall distribution of gene expression was the same across all samples, which is reasonable because they were all measured under identical conditions, and the number of genetic differences affecting gene expression levels was expected to be small (Fleischmann et al., 2002). The set of log2(Ch2/Ch1) ratios for each slide were normalized to bring the data from all slides onto the same scale as follows. (1) For each slide, the set of log ratios was median centred. (2) For each spot across all slides, the median log ratio was determined, producing a vector of spot medians. (3) For each slide, the slope of the linear regression of the log ratios from that slide versus the vector of spot medians was determined. (4) The values for each slide were scaled by dividing each log ratio of that slide by the regression slope corresponding to that slide, producing the final normalized values. This procedure resulted in consistently less variation among gene expression levels from replicate arrays of the same strain than before normalization (data available upon request).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We delineated sets of genes that were consistently expressed or unexpressed under the conditions tested. To do this, we determined the limits of detection of our microarrays. Previous experiments in our laboratory identified genomic sequences that were deleted from each of the clinical isolates compared with the H37Rv genome sequence (Kato-Maeda et al., 2001). Since there is some level of non-specific hybridization on every spot on these amplicon arrays, we used the deletion information as a negative control to model the non-specific hybridization (available online as Supplementary Figure S3 at http://mic.sgmjournals.org). Some of the isolates had no deletions, so we pooled the deletion log ratios for all strains after normalization. Because we believe that some of the log ratios of the deleted genes represent cross-hybridization, we chose the 90th percentile of these values, 0·246, as our reference cutoff value above which we believe expression is reliably detected.
After excluding those genes that were identified as variable using the Westfall & Young procedure, we defined the consistently expressed genes as those for which 90 % of the log ratios across all strains were above 0·246; 560 genes (16 % of the genes tested) fell into this category. Conversely, we defined unexpressed genes as those for which
90 % of the log ratios across all strains were below 0·246; 1382 (38 %) genes fell into this category. There were 1126 genes (31 %) that did not fall into any of the three other categories, and we called these low expression, because they fell close to the limit of detection, and therefore were within the noise level of the cutoff. Fig. 1
provides boxplots comparing the distribution of log ratios of genes in each category. The lists of genes and their expression and functional categories are available online as Supplementary Table S2 at http://mic.sgmjournals.org.
|
|
|
|
|
|
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Definition of gene expression categories
To make sure that the variability in gene expression was due to the genetic background of the strains rather than to the growth conditions, we tested each strain under exponential-phase growth in rich liquid culture. About one-sixth of the genes tested (527) displayed significantly variable expression among the strains under these conditions. We divided the remainder of the genes tested into unexpressed, expressed and low categories, with 1382, 560 and 1186 genes in these categories, respectively. We were able to assay expression levels (as opposed to relative fold changes in expression) because we co-hybridized the RNA from each strain with genomic DNA. Since each gene in the genomic DNA is present in a molar ratio of 1 compared to every other gene, it hybridizes to every spot equally except for spot-specific effects, and avoids the problem of ratios with zero in the denominator. Therefore the ratio is a normalized measurement of RNA expression level rather than a fold change. We have not calibrated the ratios to reflect the number of RNA molecules per cell, so the scale of the ratios is arbitrary, and we do not know our limits of detection. For this reason, we are unable to determine whether genes in the low category are not expressed or simply not consistently detectable in our system. A comparative analysis of the distributions of gene expression levels in eukaryotic cells (Kuznetsov et al., 2002) indicated that across a wide range of species (yeast, mouse, human), cell types and experimental conditions, the distributions are strongly skewed with many low-abundance transcripts. If prokaryotic cells behave similarly, the number of genes in each of our categories is not surprising, and it is likely that the genes in the low category are actually expressed at low levels.
Expected associations between expression categories and gene classes
Several classes of genes had unsurprising associations with the expression categories. This gave us confidence that our data and our analytical approach were valid. This is the case with the genes with functional categories insertion sequences and phages' and PE/PPE occurring more often than expected in the unexpressed and variable expression categories. It is also not surprising that the genes categorized as information pathways' occur twice as frequently as expected in the expressed category. Information pathway genes are associated with replication, transcription and translation, and include genes for sigma factors, polymerases and tRNA synthetases. These processes are expected to be highly active in bacteria undergoing exponential-phase replication. The association of genes identified as essential by Eric Rubin's group (Sassetti et al., 2003) with consistently expressed is also expected, although it should be noted that there were 82 essential genes' that were variably expressed among the clinical isolates. Finally, the association of genes deleted among a large sample of clinical isolates (Tsolaki et al., 2004
) with unexpressed and variable expression supports the idea that these genes are not essential for the growth of the organism. Only 15 of the 203 dispensable genes' in our dataset were consistently expressed.
Expression categories and functional categories
The enrichment of lipid metabolism genes in the variable category is interesting, as it is intriguing to speculate that this may have some effect on the interaction of the bacteria with the host. Lipid metabolism genes are involved in the scavenging of energy sources from the host, but they are also involved in producing the waxy coat typical of mycobacteria (Azad et al., 1997; Barry, 2001
; Cox et al., 1999
; Kolattukudy et al., 1997
; Puzo, 1990
). Lipids and genes involved in lipid metabolism have been implicated as mycobacterial virulence factors and are thought to play a role in hostpathogen interactions (Barry, 2001
; Cox et al., 1999
; Daffe & Etienne, 1999
; Hiromatsu et al., 2002
; Puzo, 1990
; Rook & Zumla, 2001
; Rosat et al., 1999
). Other studies have uncovered evidence of variations in lipid patterns among clinical isolates (Barry, 2001
; Chaicumpar et al., 1997
); it would be interesting to compare the variations in gene expression with the lipid profiles of these strains. Such a comparison may help us to home in on the specific functions of these genes, which are mostly poorly characterized.
Genes in the unknown functional category were enriched for genes in the unexpressed expression category. Genes annotated as unknown are those that have no homology with other sequences in the databases (genes of unknown function that are homologous to other genes are referred to as conserved hypotheticals). The fact that the unknowns' were less likely to be detectably expressed suggests that this functional category may contain many genes that were incorrectly identified as coding sequences. On the other hand, it is possible that these genes and homologous genes from other organisms are real genes that are unknown because they are not expressed at detectable levels under standard culture conditions (culture bias).
Expression categories and antigenic genes
We hypothesized that genes functioning as T-cell antigens might be more likely to be variably expressed among clinical isolates, as a result of immune selection. Although the numbers were small, there was a significant enrichment for variable and consistently expressed genes among the genes identified as T-cell antigens by Belisle's group (Covert et al., 2001). The consistently expressed gene enrichment is not surprising: since Covert et al. used a proteomics approach to identify the T-cell antigens, it is likely that their results were biased for highly expressed genes. It is interesting, however, that the T-cell antigen list was enriched for genes that are variably expressed among the clinical isolates. This suggests that these genes might be undergoing immune selection in vivo, resulting in changes in their expression. While the variability in expression of these genes supports the idea that they are antigenic, it is not promising for those who would want to use them as vaccine antigens, as they may not be expressed at high levels in all strains.
In a similar vein, we were interested in whether or not there was evidence in our dataset that PE and PPE genes might be variably expressed among the strains tested. These mysterious gene families are unique to mycobacteria and were discovered when the H37Rv genome was sequenced (Cole et al., 1998). It has been hypothesized that these genes may contribute to antigenic diversity and act as a sort of decoy for the immune system. There is evidence that some of these genes can be expressed at the cell surface, are antigenic, and are variable among clinical isolates (Banu et al., 2002
; Brennan & Delogu, 2002
; Brennan et al., 2001
; Sampson et al., 2001
). Because these genes are members of families of highly related proteins, it has been difficult to determine whether clinical isolates are expressing different members of these families, since antibodies often cross-react with multiple family members, and the repetitive sequences make the RNA difficult to amplify and sequence. For this reason, we used an informatics approach to identify spots that were likely to be measuring unique sequences of the genes. After excluding spots that were likely to cross-hybridize with other members of the gene families, 77 of the 112 PE and PPE genes represented in our dataset remained. Most of these 77 genes (54) were expressed at low to undetectable levels, but 20 of them were variably expressed among the clinical isolates. Only three of them were consistently expressed. Therefore our data support the idea that members of these gene families are polymorphic and expressed differentially among clinical isolates.
Expression categories and common vaccine, diagnostic and drug target candidate genes
Three major vaccine and diagnostic targets, esat-6, Antigen 85A and 19 kDa antigen, were classified as variable in our system. Although esat-6 was designated variable among strains, it was expressed at high (albeit variable) levels in all strains. Therefore our results do not call into question its viability as a vaccine antigen that should be effective against many strains of M. tuberculosis. Similarly, Antigen 85A, while expressed on average at much lower levels than esat-6, was still well within the range of detectable expression in all strains. The 19 kDa antigen was expressed overall at lower levels than Antigen 85A, and in some instances hovered near the limit of detectable expression in our system. If high expression levels are important for vaccine or diagnostic targets, the 19 kDa antigen might be a poor candidate molecule. However, if low levels of expression are sufficient, it would be of interest to test this candidate using a more sensitive method, such as RT-PCR.
Drug targets are a bit trickier to assess, since protein concentration within a cell may affect drug efficacy. (It is important to remember that mRNA levels may not always correlate with protein concentrations.) Furthermore, the conditions under which a gene is expressed are important in developing therapeutics. There is considerable interest in identifying drug targets that are expressed during the persistent stage of infection, such as isocitrate lyase and pcaA (Smith et al., 2004). Not surprisingly, isocitrate lyase is expressed at low to undetectable levels in all strains tested under the exponential-phase rich growth conditions we used. On the other hand, pcaA, a gene involved in mycolic acid synthesis, and which has been implicated in mycobacterial persistence in mice, is consistently expressed in our system. Another interesting candidate drug target is nrdF2 (Kana & Mizrahi, 2004
), a gene encoding the small subunit of ribonucleotide reductase. This gene has been shown by Kana & Mizrahi (2004)
to be essential under normal culture conditions and is also consistently expressed in our system. Because of the difficulties in interpreting expression data, especially data collected under exponential-phase growth conditions in liquid culture, we recommend using caution when extrapolating our results for potential drug targets. However, our results do point out that strain variation in expression does exist, so genes of interest should be assayed in several strains under relevant conditions before proceeding with drug development.
Conclusions
Variation in gene expression among clinical isolates has implications for pathogenicity and the identification of candidate genes for drug targets, vaccine antigens and diagnostic assays. The enrichment of lipid metabolism and PE/PPE genes and T-cell antigens for genes that are variably expressed suggests that clinical isolates may differ in their host interactions. Variable expression of some genes thought to be essential for bacterial survival highlights the importance of considering strain-to-strain variation when selecting candidate targets for therapeutics and diagnostics. We propose that, as techniques become optimized for studying global gene expression profiles of M. tuberculosis under conditions that are more closely related to host infection, strain-to-strain gene expression diversity should be examined under these conditions. Since such experiments are currently quite difficult and labour-intensive to perform, any genes of interest should be examined individually for gene expression variability among strains under relevant conditions.
![]() |
ACKNOWLEDGEMENTS |
---|
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Banu, S., Honore, N., Saint-Joanis, B., Philpott, D., Prevost, M. C. & Cole, S. T. (2002). Are the PE-PGRS proteins of Mycobacterium tuberculosis variable surface antigens? Mol Microbiol 44, 919.[CrossRef][Medline]
Barry, C. E., 3rd (2001). Interpreting cell wall virulence factors' of Mycobacterium tuberculosis. Trends Microbiol 9, 237241.[CrossRef][Medline]
Behr, M. A., Wilson, M. A., Gill, W. P., Salamon, H., Schoolnik, G. K., Rane, S. & Small, P. M. (1999). Comparative genomics of BCG vaccines by whole-genome DNA microarray. Science 284, 15201523.
Betts, J. C., Lukey, P. T., Robb, L. C., McAdam, R. A. & Duncan, K. (2002). Evaluation of a nutrient starvation model of Mycobacterium tuberculosis persistence by gene and protein expression profiling. Mol Microbiol 43, 717731.[CrossRef][Medline]
Brennan, M. J. & Delogu, G. (2002). The PE multigene family: a molecular mantra for mycobacteria. Trends Microbiol 10, 246249.[CrossRef][Medline]
Brennan, M. J., Delogu, G., Chen, Y., Bardarov, S., Kriakov, J., Alavi, M. & Jacobs, W. R., Jr (2001). Evidence that mycobacterial PE_PGRS proteins are cell surface constituents that influence interactions with other cells. Infect Immun 69, 73267333.
Chaicumpar, K., Fujiwara, N., Nishimura, O., Hotta, H., Pan, J. W., Takahashi, M., Abe, C. & Yano, I. (1997). Studies of polymorphic DNA fingerprinting and lipid pattern of Mycobacterium tuberculosis patient isolates in Japan. Microbiol Immunol 41, 107119.[Medline]
Cole, S. T., Brosch, R., Parkhill, J. & 39 other authors (1998). Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393, 537544.[CrossRef][Medline]
Covert, B. A., Spencer, J. S., Orme, I. M. & Belisle, J. T. (2001). The application of proteomics in defining the T cell antigens of Mycobacterium tuberculosis. Proteomics 1, 574586.[CrossRef][Medline]
Cox, J. S., Chen, B., McNeil, M. & Jacobs, W. R., Jr (1999). Complex lipid determines tissue-specific replication of Mycobacterium tuberculosis in mice. Nature 402, 7983.[CrossRef][Medline]
Daffe, M. & Etienne, G. (1999). The capsule of Mycobacterium tuberculosis and its implications for pathogenicity. Tuber Lung Dis 79, 153169.[CrossRef][Medline]
Dudoit, S., Yang, Y. H., Callow, M. J. & Speed, T. P. (2002). Statistical methods for identifying differentially expressed genes in replicated microarray experiments. Statistica Sinica 12, 111139.
Dudoit, S., Shaffer, J. P. & Boldrick, J. C. (2003). Multiple hypothesis testing in microarray experiments. Stat Sci 18, 71103.[CrossRef]
Fellenberg, K., Hauser, N. C., Brors, B., Neutzner, A., Hoheisel, J. D. & Vingron, M. (2001). Correspondence analysis applied to microarray data. Proc Natl Acad Sci U S A 98, 1078110786.
Fisher, M. A., Plikaytis, B. B. & Shinnick, T. M. (2002). Microarray analysis of the Mycobacterium tuberculosis transcriptional response to the acidic conditions found in phagosomes. J Bacteriol 184, 40254032.
Fleischmann, R. D., Alland, D., Eisen, J. A. & 23 other authors (2002). Whole-genome comparison of Mycobacterium tuberculosis clinical and laboratory strains. J Bacteriol 184, 54795490.
Girke, T., Todd, J., Ruuska, S., White, J., Benning, C. & Ohlrogge, J. (2000). Microarray analysis of developing Arabidopsis seeds. Plant Physiol 124, 15701581.
Greenacre, M. (1984). Theory and Applications of Correspondence Analysis. London: Academic Press.
Hiromatsu, K., Dascher, C. C., LeClair, K. P., Sugita, M., Furlong, S. T., Brenner, M. B. & Porcelli, S. A. (2002). Induction of CD1-restricted immune responses in guinea pigs by immunization with mycobacterial lipid antigens. J Immunol 169, 330339.
Jasmer, R. M., Hahn, J. A., Small, P. M. & 7 other authors (1999). A molecular epidemiologic analysis of tuberculosis trends in San Francisco, 19911997. Ann Intern Med 130, 971978.
Kana, B. D. & Mizrahi, V. (2004). Molecular genetics of Mycobacterium tuberculosis in relation to the discovery of novel drugs and vaccines. Tuberculosis (Edinb) 84, 6375.[CrossRef][Medline]
Kane, M. D., Jatkoe, T. A., Stumpf, C. R., Lu, J., Thomas, J. D. & Madore, S. J. (2000). Assessment of the sensitivity and specificity of oligonucleotide (50mer) microarrays. Nucleic Acids Res 28, 45524557.
Kato-Maeda, M., Rhee, J. T., Gingeras, T. R., Salamon, H., Drenkow, J., Smittipat, N. & Small, P. M. (2001). Comparing genomes within the species Mycobacterium tuberculosis. Genome Res 11, 547554.
Kolattukudy, P. E., Fernandes, N. D., Azad, A. K., Fitzmaurice, A. M. & Sirakova, T. D. (1997). Biochemistry and molecular genetics of cell-wall lipid biosynthesis in mycobacteria. Mol Microbiol 24, 263270.[Medline]
Kuznetsov, V. A., Knott, G. D. & Bonner, R. F. (2002). General statistics of stochastic process of gene expression in eukaryotic cells. Genetics 161, 13211332.
Manganelli, R., Voskuil, M. I., Schoolnik, G. K., Dubnau, E., Gomez, M. & Smith, I. (2002). Role of the extracytoplasmic-function sigma factor sigma(H) in Mycobacterium tuberculosis global gene expression. Mol Microbiol 45, 365374.[CrossRef][Medline]
Puzo, G. (1990). The carbohydrate- and lipid-containing cell wall of mycobacteria, phenolic glycolipids: structure and immunological properties. Crit Rev Microbiol 17, 305327.[Medline]
Rook, G. A. & Zumla, A. (2001). Advances in the immunopathogenesis of pulmonary tuberculosis. Curr Opin Pulm Med 7, 116123.[CrossRef][Medline]
Rosat, J. P., Grant, E. P., Beckman, E. M. & 7 other authors (1999). CD1-restricted microbial lipid antigen-specific recognition found in the CD8+ alpha beta T cell pool. J Immunol 162, 366371.
Sampson, S. L., Lukey, P., Warren, R. M., van Helden, P. D., Richardson, M. & Everett, M. J. (2001). Expression, characterization and subcellular localization of the Mycobacterium tuberculosis PPE gene Rv1917c. Tuberculosis (Edinb) 81, 305317.[CrossRef][Medline]
Sassetti, C. M., Boyd, D. H. & Rubin, E. J. (2003). Genes required for mycobacterial growth defined by high density mutagenesis. Mol Microbiol 48, 7784.[CrossRef][Medline]
Smith, C. V., Sharma, V. & Sacchettini, J. C. (2004). TB drug discovery: addressing issues of persistence and resistance. Tuberculosis (Edinb) 84, 4555.[CrossRef][Medline]
Talaat, A. M., Howard, S. T., Hale, W. T., Lyons, R., Garner, H. & Johnston, S. A. (2002). Genomic DNA standards for gene expression profiling in Mycobacterium tuberculosis. Nucleic Acids Res 30, e104.
Tsolaki, A. G., Hirsh, A. E., DeRiemer, K. & 7 other authors (2004). Functional and evolutionary genomics of Mycobacterium tuberculosis: insights from genomic deletions in 100 strains. Proc Natl Acad Sci U S A 101, 48654870.
Valway, S. E., Sanchez, M. P., Shinnick, T. F., Orme, I., Agerton, T., Hoy, D., Jones, J. S., Westmoreland, H. & Onorato, I. M. (1998). An outbreak involving extensive transmission of a virulent strain of Mycobacterium tuberculosis. N Engl J Med 338, 633639.
Wilson, M., DeRisi, J., Kristensen, H. H., Imboden, P., Rane, S., Brown, P. O. & Schoolnik, G. K. (1999). Exploring drug-induced alterations in gene expression in Mycobacterium tuberculosis by microarray hybridization. Proc Natl Acad Sci U S A 96, 1283312838.
Xu, W., Bak, S., Decker, A., Paquette, S. M., Feyereisen, R. & Galbraith, D. W. (2001). Microarray-based analysis of gene expression in very large gene families: the cytochrome P450 gene superfamily of Arabidopsis thaliana. Gene 272, 6174.[CrossRef][Medline]
Received 3 August 2004;
revised 31 August 2004;
accepted 2 September 2004.
HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
INT J SYST EVOL MICROBIOL | MICROBIOLOGY | J GEN VIROL |
J MED MICROBIOL | ALL SGM JOURNALS |