Pii: s1369-5274(02)00280-10

Functional and comparative genomics of pathogenic bacteria
Gary K Schoolnik
Microarray expression profiling and the development of data- needed to identify significantly regulated genes [14••,15•,16•].
mining tools and new statistical instruments affords an Although not the topic of this review, some of the most unprecedented opportunity for the genome-scale study of comprehensive microarray studies to date concern the bacterial pathogenicity. Expression profiles obtained from bacteria non-pathogenic model systems Escherichia coli K-12 [17–23], grown in media simulating host microenvironments yield a portrait Bacillus subtilis [24] and Caulobacter crescentus [25]. of interacting metabolic pathways and multistage developmentalprograms and disclose regulatory networks. The analysis of This review focuses on two aspects of bacterial pathogenicity, closely related strains and species by microarray-based gleaned from an emerging literature that describes the use comparative genomics provides a measure of genetic variability of genome microarrays to study the biology of infectious within natural populations and identifies crucial differences agents. Included here are expression profiling studies of between pathogen and commensal. In the near future, the bacteria cultured in vitro under conditions intended to combined use of bacterial and host microarrays to study the induce in vivo transcriptional programs, to define regulon same infected tissue will reveal the host–pathogen dialogue in a membership, and to illuminate crucial biosynthetic pathways.
gene-by-gene and site- and time-specific manner. This review Also discussed are the capacity of comparative genomic discusses the use of microarray-based expression profiling to studies by microarray methods to characterize the genetic identify genes of pathogenic bacteria that are differentially variability of natural populations, and to identify differences regulated in response to host-specific signals. Additionally, the between pathogen and commensal, and between related review describes the application of microarray methods to pathogens that occupy different host niches or vary in some disclose differences in gene content between taxonomically other subtle manner. The review concludes with a brief related strains that vary with respect to pathogenic phenotype.
summary of expression studies of infected host cells andthe promise that this holds for capturing the dialogue Addresses
Beckman Center, Room 241, Stanford Medical School, Stanford,California 94305, USA; e-mail: schoolni@cmgm.stanford.edu Microarray expression profiling of pathogenic
Current Opinion in Microbiology 2002, 5:20–26
bacteria
The ultimate goal of whole genome expression studies of
pathogenic bacteria is the identification of bacterial genes 2002 Elsevier Science Ltd. All rights reserved.
that are differentially regulated in the host. Within this class Abbreviations
of genes are those that adapt the microbe to host-specific microenvironments or encode virulence determinants.
Ideally, studies of this kind would compare expression profiles of bacteria within infected tissues with profiles from bacteria cultured under standardized in vitro conditions ofgrowth. Unfortunately, this technically formidable goal has Introduction
not been achieved using non-amplification methods because A microarray is a device that provides a surface containing the number of organisms within infected tissues is often representations of all (or most) of the identified open small, RNA from host cells is vastly more abundant than reading frames (ORFs) of a sequenced and annotated bacterial RNA, and no efficient method to differentially genome. Whether fabricated as a glass-spotted microarray extract stabilized bacterial RNA from tissues has been [1–3], high-density oligonucleotide array [4,5] or nylon described. Because of these considerations, all published membrane macroarrray, this simple experimental system microarray reports of pathogenic bacteria have studied provides the basis for two quite distinctive experimental organisms grown in vitro. Inherent in this experimental paradigms. Functional genomics uses expression profiling design is the quite reasonable assumption that the expression of mRNA to provide a condition-specific and time- of genes encoding virulence determinants can often be induced specific genome-scale snapshot of the transcriptome [6,7].
in vitro by simple modifications of laboratory media [26].
Comparative genomics contrasts two or more genomes atthe ORF-content level of resolution [8,9]. Both applications Iron limitation
entail the use of clustering algorithms [10,11] and pathway Paustian et al. [27•] studied the transcriptional response of databases [12,13] to identify co-regulated genes that the animal pathogen Pasteurella multocida growing under perform common metabolic and biosynthetic functions.
iron-limited conditions in vitro. Available free iron is limited However, microarray expression profiling data poses special in host tissues and iron deprivation is known to induce a analytical challenges that have required the development complex iron-scavenging response. To characterize the of new statistical instruments and the recognition that P. multocida low-iron response, the organism was grown multiple biological replicates of the same experiment are either in an iron-replete medium or in the same medium Functional and comparative genomics of pathogenic bacteria Schoolnik 21
containing an iron chelator. The expression profile, acidic environment. To identify the transcriptional obtained with a DNA microarray containing 96% of the response of H. pylori to acidic conditions of growth, Ang et al.
organism’s 2014 ORFs, comprised 135 genes (~7% of the [31] grew a recent gastric isolate on Columbia agar titrated genome), including many known in other organisms to be either to pH 7.2 or to pH 5.5. An expression profile iron-regulated. However, the differential regulation of obtained 48 hours thereafter with a membrane macroarray many other genes showed that the transcriptional response containing 96% of the 1534 predicted ORFs of H. pylori to a change in the concentration of only one metal is strain 26695 identified 80 acid-upregulated ORFs. Sixteen complex and pleiotrophic, and points to the interdependency ORFs were previously known to be acid-induced either of multiple metabolic pathways. This result also raises the in H. pylori or in other species, whereas an additional 43 following question, typical of many whole genome functionally annotated ORFs code for proteins not pre- expression responses: which genes in the overall expression viously known to be acid-regulated. Thus, only a minority profile compose the primary or direct response to iron of the identified acid-regulated genes could be directly limitation, and which are secondary, downstream conse- associated with acid tolerance by a plausible physiological quences of an intracellular deficiency in iron that might mechanism or by reference to prior work. In part, this affect multiple pathways? Direct effects are best identified could reflect the complex and relatively unexplored nature by the use of short time courses and experimental conditions of the acid-resistant phenotype in this species and the that do not cause differences in growth rate between the large number of its ORFs that code for proteins of unknown function. However, in part, this may also comefrom this study’s use of a solid medium, which could result Adaptation to acidic microenvironments
in a heterogeneous growth environment because of the The pH of some host microenvironments, like the con- generation of metabolic gradients in the surrounding agar.
centration of available iron, challenges the capacity of It may also result from the protracted 48-hour time course, bacteria to adapt, especially in strongly acidic organs. The at the end of which the transcriptome likely reflects a new most extreme of these, the stomach, normally reaches pH steady state rather than the adaptive process itself. values as low as 2.0, and virtually all bacteria that ultimatelycompose the rich and varied microflora of the colon and Cell-density-dependent gene regulation
distal small intestine have successfully transited the Streptococcus pneumoniae, a common cause of pneumonia ‘gastric acidity barrier’ before coming to reside in the more and meningitis, ordinarily colonizes pharyngeal mucous alkaline environment of the intestine. Some pathogenic membranes, a site where the development of genetic species, particularly Shigella spp. and the enterohemorrhagic diversity in this species is favored by transformation with and enteroinvasive E. coli biotypes are particularly acid- DNA from the other strains and species that inhabit this resistant [28], explaining why infections with these ecosystem. Competence — the capacity to bind and take up organisms can be initiated by fewer than 30 bacteria.
extraneous DNA — is a transient, temporally programmedphysiological state induced by competence-stimulating Growth of E. coli with acetate, which lowers the intracellular peptide (CSP). Rimini et al. [32••] studied the kinetics of pH, significantly increases the acid resistance of the E. coli.
CSP-induced gene expression using a S. pneumoniae To learn more about the mechanism of acetate-induced acid membrane macroarray. After addition of CSP, results from tolerance, Arnold et al. [29•] incubated enterohemorrhagic expression profiling corroborated the kinetics of the E. coli serotype O157:H7 in a pH 7.0 medium, with or previously documented early and late competence without 100 mM acetate, and then obtained an expression transcriptional patterns. However, the identification of profile using an E. coli K-12 genome membrane macroarray 23 other upregulated genes not previously recognized to be lacking the 1387 ORFs that are present in the E. coli associated with competence was more important, as was the O157:H7 genome but absent in the genome of E. coli K-12 unexpected discovery of seven downregulated CSP genes.
[30]. Despite this limitation, among the 26 upregulated This study demonstrates how microarray expression genes were six genes previously known to specify functions profiling can illuminate a programmed physiological that defend the intracellular pH during acidic conditions of process by relating the transcriptional state of genes to a growth. Also induced by acetate were: cfa, whose product time course that marks its initiation, manifestation and cyclopropanates unsaturated fatty acids in the inner decline. It also shows that a genome-wide study will nearly membrane, possibly decreasing its permeability to protons; always reveal new information, even about a well-studied hdeA, which specifies a periplasmic chaperon hypothesized phenomenon. Other examples of this kind with non- to prevent acid-induced denaturation of periplasmic pathogenic bacteria include studies of the cell cycle in proteins; and three oxidative-stress genes, dps, katE and C. crescentus and sporulation in B. subtilis [24,25].
grxB, indicating that exposure to acetate generates reactiveoxygen species.
De Saizieu et al. [33••] used a high-density oligonucleotideAffymetrix S. pneumoniae array to study a bacteriocin-like Helicobacter pylori is the most remarkable of the acid tolerant peptide (BlpC) two-component quorum-sensing system group. It chronically colonizes gastric mucous membranes that is remarkably similar with respect to its regulation, and is, thus, normally exposed, often for decades, to an processing, export and signal transduction to the competence Host–microbe interactions: bacteria
system described above. The expression profile that was factors are of particular interest because, in other species, induced by the addition of BlpC to an exponentially some have been found to confer adaptive responses to growing culture included 16 genes that are clustered on environmental factors and stress or to be required for the chromosome near bfpHR, which encodes the cognate virulence. Manganelli et al. [38••] used a combination of two-component system. These results and the demonstration mutational and microarray methods to define the function by Throup et al. [34] that a blpHR mutant exhibits attenuated and characterize the regulon of the M. tuberculosis ECF σ virulence in a murine model of pneumococcal pneumonia factor, σE. Disruption of sigE yielded a strain that was more shows that microarray expression analysis was able to sensitive than the wild-type parent to heat shock, the ionic identify a quorum-sensing system that is expressed in vivo detergent sodium dodecyl sulfate (SDS) and to oxidants, and that exhibited impaired growth in macrophage celllines. To identify σE-regulated genes, expression profiles Low-oxygen gene regulation and induction of dormancy
were obtained from mid-exponential-phase cultures of a Like S. pneumoniae, Mycobacterium tuberculosis mainly infects σE mutant and the wild-type parent that had been exposed the lung. However, unlike patients with pneumococcal to 0.05% SDS, a treatment that likely perturbs cell pneumonia, an acute infectious process, most people envelope lipids. Twenty-three genes were identified infected with M. tuberculosis have a latent form of the whose SDS-induced expression required σE. Among these disease and are non-infectious and asymptomatic. In vitro, were sigB, aceA (which encodes isocitrate lyase of the a state of non-replicating persistence can be produced by glyoxalate shunt, an activity required for full virulence in a allowing a non-stirred culture to generate a low oxygen murine model of tuberculosis) [39] and fadB2 (which gradient as the respiring bacteria settle to the bottom of a encodes 3-hydroxyacyl CoA dehydrogenase, a component culture tube [35]. Non-replicating cultures of this kind, of the fatty-acid β-oxidation pathway that may enable viable for months and perhaps years, can be resuscitated to assimilation of host fatty acids). These data and results from the replicating state by re-introduction of oxygen. To the preceding study demonstrate that microarray expression identify M. tuberculosis genes differentially regulated by analysis can, in principle, identify gene sets whose regulation hypoxia, Sherman et al. [36••] shifted an early exponential- requires, directly or indirectly, each of the annotated phase culture from growth in air (~20% O2) to growth in a alternative σ factors or two-component regulators in the hypoxic atmosphere (0.2% O2, 99.8% N2). The genes genome. Moreover, because some of the regulated genes will induced by hypoxia were identified using a DNA be other transcription factors, it should be possible to recon- microarray containing >97% of the 3924 identified ORFs.
struct hierarchies of regulatory networks from these data.
Forty-seven ORFs were upregulated and, althoughapproximately two-thirds of the induced genes are of Inhibition of biosynthetic pathways and the identification
unknown function, several with annotated functions are of new drug targets
plausibly involved with adaptation to hypoxia. Among these M. tuberculosis is not only a formidable human pathogen, genes is acr, which encodes α-crystallin, a 14 kDa protein but is also increasingly difficult to treat because of with chaperone activity that was previously shown to emerging resistance to one or more antitubercular drugs, accumulate during non-replicating persistence [37]. Several including isoniazid (INH). It has long been known that genes coding for putative transcription factors were also INH blocks the biosynthesis of mycolic acids, an essential identified within the induced gene set, including two component of the mycobacterial cell envelope, and recent within a three-gene operon. One of these two is predicted biochemical studies indicate that it does so by inhibiting to encode a membrane-bound sensor histidine kinase and the type II fatty-acid synthase (FAS-II) complex that is the other a two-component response regulator. Mutational required for the production of the full-length meromycolate analysis of this operon showed that disruption of the chain, either by binding NADH within the active site of response regulator, but not the adjacent sensor histidine enoyl-acyl carrier protein reductase (InhA) [40] or by forming kinase, prevented hypoxic induction of acr, whose a ternary complex with β-ketoacyl-ACP-synthase (KasA) expression was thought to reflect the transcriptional state of and the acyl carrier protein, AcpM [41].
other genes within the low-oxygen stimulon. Compared tothe wild-type parent strain, this mutant survived less well Wilson et al. [42] obtained expression profiles from in late stationary phase. The identification of the cognate INH-treated mid-log phase cultures of M. tuberculosis; only response regulator of this low-oxygen response exemplifies 14 differentially regulated ORFs were identified out of the how microarray expression analysis can be used to identify 3834 whose expression state had been modified. Amongst transcription factors and dissect regulatory networks.
these genes was the induction by INH of an operon-likecluster that encodes components of the FAS-II complex, σ factor regulator cascades
including AcpM and KasA. This result, evident after only M. tuberculosis dormancy may be considered to be a kind of 40 minutes, indicates that an expression profile can developmental program that is characterized by temporally provide useful information about a compound’s mode of ordered transcriptional events governed by a hierarchy of action and demonstrates that inhibition of a biosynthetic transcription factors, including alternative σ factors. The pathway is sensed and responded to at the transcriptional extracytoplasmic function (ECF) subset of alternative σ level within minutes. Also induced was fbpC, which Functional and comparative genomics of pathogenic bacteria Schoolnik 23
encodes trehalose-dimycolyl transferase, an activity at the of the tested pathogenic M. bovis strains. Furthermore, end of the mycolate biosynthetic pathway that esterifies compared to pathogenic M. bovis strains, five additional mycolic acids with cell-wall carbohydrates. This result regions containing 38 ORFs were deleted from one or more shows that an expression result can illuminate components of the tested BCG strains. Analysis of the regions deleted of a multicomponent pathway that are remote from its from BCG, but present in the sequenced M. tuberculosis direct site of action — newly identified pathway components strain, showed that genes classified as transcriptional could contribute to the drug discovery process by disclosing regulators were lost disproportionately and may, thus, control the expression of genes required for virulence.
When analyzed within a historical context, these results Microarray-based comparative genomics
show that microarray-based comparative genomics can be Genetic variability and natural selection yield strains and used to reconstruct the genealogy of related strains at the species adapted to particular microenvironments of the host genomic level of resolution. In a second study by the same and result in phenotypic differences between non- group, Kato-Maeda et al. [46•] used a high-density oligonu- pathogenic commensals and virulent biotypes. Accordingly, cleotide Affymetrix array to compare 19 recent clinical genomic comparisons between pathogenic and non- isolates of M. tuberculosis. Compared to the sequenced pathogenic strains of the same species can be particularly reference strain, each unique clinical isolate was found to informative because genes exclusively present in the former have lost, on average, ~17 ORFs corresponding to ~0.3% of may be required for infectivity, virulence or adaptation to a the H37Rv genome. In all, 25 deleted sequences were particular host niche. Microarray-based comparisons detected, including 22 intergenic segments and all or part between a fully sequenced genome and an unsequenced, of 93 ORFs. On the basis of their functional annotations, but related, genome can provide valuable information several of the deleted genes could conceivably affect about the diversity and evolution of pathogens and virulence, including three encoding phospholipase-C, one symbionts [8,9]. Comparisons of this kind employ a encoding a polyketide synthase and three encoding microarray containing representations of all the ORFs of putative transcriptional regulators. Remarkably, strains the sequenced, reference strain and labeled DNA from that had sustained the most deletions were less likely to the unsequenced, experimental strain. The resulting have been isolated from patients with pulmonary cavitation.
hybridized array will disclose genes common to both strains Cavity formation is a hallmark of tuberculosis and essential and genes that are present in the reference strain but absent for the efficient transmission of the organism to susceptible in the experimental strain. This method, however, cannot hosts. Thus, degradation of the genome may be associated detect genes present in the experimental strain but absent in this species with a trend to decreased infectivity.
in the reference strain: point mutations, includingframe-shift mutations; small deletions and deletions in H. pylori strain diversity
homologous repetitive elements; rearrangements of the H. pylori infection of the upper gastrointestinal tract genome that have not resulted in deletion of a gene; and causes a spectrum of conditions ranging from asymptomatic differences in the number of multicopy genes [8,9,43••].
infection to gastritis, gastric and duodenal peptic ulcer Additionally, in contrast to high-density oligonucleotide disease and gastric cancer. Comparison of two complete arrays, DNA-spotted microarrays and membrane macroarrays H. pylori sequences revealed that ~6% of each genome was do not ordinarily include representations of intergenic not present in the other genome and that recombinations, regions of the genome and, thus, cannot detect deletions insertions and deletions, changes in repetitive elements within these non-coding segments, even though these and single-nucleotide substitutions had created considerable specify promoter elements and small, non-translated RNAs diversity [47]. To further explore the genomic diversity of and thus could be functionally important [44]. Despite this species, Salama et al. [43••] used a microarray representing these limitations, the few published studies of this kind 98.6% of the ORFs of both sequenced species to examine have been quite informative, in part because events leading the genomic content of 15 H. pylori strains. They identified to gene acquisition and gene loss are a major source of 1281 ORFs that were common to all the tested strains; diversity in bacterial pathogens [8] and many changes of these represent the ‘functional core’ of this species’ this kind are readily detected by microarray methods.
genome. Among them were genes coding for metabolicand biosynthetic pathways and for cellular and regulatory M. tuberculosis, M. bovis and BCG vaccine strains
functions. By contrast, 362 ORFs, comprising 22% of the Behr et al. [45], in perhaps the first example of a study of genome, were absent from one or more of the tested this kind, used a DNA microarray to compare the genome strains; these comprise strain-specific genes and were composition of the sequenced M. tuberculosis laboratory hypothesized to encode functions that adapt the organism strain H37Rv with the closely related pathogenic to a particular host niche. An intriguing aspect of this study species, M. bovis, and with several strains of the bacille was the use of a clustering algorithm for the analysis of Calmette–Guerin (BCG) vaccine variant that was produced strain-specific genes and the identification of several genes by serial in vitro passage of M. bovis between 1908 and that may have been co-inherited with genes in the patho- 1921. Compared to M. tuberculosis, 11 regions containing 91 genicity island and may therefore also encode virulence ORFs were found to have been deleted from one or more determinants. In a separate study, Israel et al. [48] used the Host–microbe interactions: bacteria
same H. pylori microarray to compare the genomic content expression studies of infected host cells have been of two clinical strains that produce significantly different conducted as well and hold considerable promise for levels of gastritis, cellular proliferation and apoptosis in the capturing the intricate sequence of measure and counter- gerbil gastritis model. The microarray results showed that measure between pathogen and host. Thus far, however, the less proinflammatory strain had sustained a large all such studies have been carried out in vitro. Technical deletion of the cag pathogenicity island, providing a innovations will be necessary before microarray-based genetic explanation for its relative attenuation.
bacterial expression studies of infected tissues can beconducted. The availability of improved methods and more Profiling the dialogue between pathogen
powerful bioinformatic tools will provide whole-genome portraits of the transcriptomes of pathogen and host in a Pathogenesis entails not only the differential expression of time-, tissue- and cell-specific manner.
bacterial genes, but also responses by the host. In principle,then, microarray expression analysis of bacterially infected Acknowledgements
cells and tissues can identify, simultaneously and in the I thank past and current members of my laboratory for their contributions tothe art and science of microarray expression work in my laboratory, which same sample, host and pathogen genes that are regulated was supported by grants from the National Institutes of Health (AI 39521, during the infectious process. Although not within the AI 43422 and AI44826) and the Glaxo Group Research Action TB Program.
scope of this review, the host contribution to this processhas been the focus of five published microarray studies References and recommended reading
[49•,50••,51•,52•,53••] characterizing the host-cell response Papers of particular interest, published within the annual period of review,have been highlighted as: to attached or invading bacteria. Each of these studies usedtransformed cell lines, relatively large multiplicities-of- infection (MOIs), and short time courses, so the results Eisen MB, Brown PO: DNA arrays for analysis of gene expression.
reflect early events. Three of these studies employed Methods Enzymol 1999, 303:179-205.
macrophage cell lines [49•,50••,51•] and therefore explored De Risi JL, Iyer VR, Brown PO: Exploring the metabolic and genetic
aspects of innate rather than acquired immunity. Four of control of gene expression on a genomic scale. Science 1997,
these studies [49•,50••,52•,53••] also obtained expression 278:680-686.
profiles of cell lines exposed to purified bacterial products, Schoolnik GK, Voskuil MI, Schnappinger D, Yildiz FH, Meibom K,
Dolganov NA, Wilson MA, Chong KH: Whole genome DNA
including pro-inflammatory cell-wall constituents and toxins, microarray expression analysis of biofilm development by Vibrio
or to bacteria with mutations in genes coding for virulence cholerae O1 El Tor. Methods Enzymol 2001, 336:3-18.
determinants. This successful strategy identified microbial Lipshutz RJ, Fodor SP, Gingeras TR, Lockhart DJ: High density
molecules responsible for the induction of some components synthetic oligonucleotide arrays. Nat Genet 1999, 21(Suppl 1):20-24.
of the host-cell transcriptional response.
Lockhart DJ, Byrne MC, Follettie MT, Gallo MV, Chee MS, Mittmann M,
Wang C, Kobayashi M, Horton H, Brown EL: Expression monitoring
by hybridization to high-density oligonucleotide arrays. Nat
The referenced studies illustrate that experiments using Biotechnol 1996, 14:1675-1680.
transformed cell lines are informative, but results from the Ferea TL, Brown PO: Observing the living genome. Curr Opin Gen
use of primary cells freshly isolated from tissues, for Dev 1999, 9:715-722.
example, primary bone-marrow-derived macrophages, may Harrington CA, Rosenow C, Retief J: Monitoring gene expression
better simulate in vivo conditions. However, studies using using DNA microarrays. Curr Opin Microbiol 2000, 3:285-291.
single cell types, whether primary or transformed, lack the Ochman H, and Jones IB: Evolutionary dynamics of full genome
content in Escherichia coli. EMBO 2000, 19:6637-6643.
context of a multicellular milieu where signaling betweencells of different lineages modulates the response of Ochman H, Moran NA: Genes lost and genes found: evolution of
bacterial pathogenesis and symbiosis. Science 2001,
individual cells to an infectious agent. This kind of 292:1096-1098.
complexity is difficult to simulate in vitro and will require 10. Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and
technical innovations that permit expression profiling of display of genome-wide expression patterns. Proc Natl Acad Sci
USA 1998, 95:14863-14868.
individual cell types within infected tissues.
11. Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E, Lander ES, Golub TR: Interpreting patterns of gene expression with
Conclusions
self-organizing maps: methods and application to hematopoietic
The studies reviewed here show that microarray expression differentiation. Proc Natl Acad Sci USA 1999, 96:2907-2912.
profiling is a powerful method to identify genes differentially 12. Karp PD, Krummenacker M, Paley S, Wagg J: Integrated
pathway-genome databases and their role in drug discovery.
regulated by biochemical signatures of host microenviron- Trends Biotechnol 1999, 17:275-281.
ments, genes that are controlled, directly or indirectly, by 13. Karp PD, Riley M, Saier M, Paulsen IT, Paley SM, Pellegrini-Toole A: transcription factors and genes that code for components of The EcoCyc and MetaCyc databases. Nucleic Acids Res 2000,
multistep metabolic and biosynthetic pathways. Equally 28:56-59.
illuminating are microarray-based comparative studies to 14. Arfin SM, Long AD, Ito ET, Tolleri L, Riehle MM, Paegle ES, assess the extent and nature of genetic variability within Hatfield GW: Global gene expression profiling in Escherichia coli
K-12. J Biol Chem 2000, 275:29672-29684.
natural populations of related species and strains and to This is a seminal study that proposes a new statistical method for the delineate differences, at the ORF level of resolution, analysis of microarray expression data on the basis of a linear analysis ofvariance (ANOVA) model. Use of this instrument to analyze an E. coli K-12 between pathogen and commensal. Complementary membrane macroarray study of the effect of integration host factor showed Functional and comparative genomics of pathogenic bacteria Schoolnik 25
that arbitrarily selected fold-difference criteria cannot reliably identify study in genomic scale by microarray. Infect Immun 2001,
genes whose expression levels are significantly different between two 69:1679-1686.
32. Rimini R, Jansson B, Feger G, Roberts TC, de Francesco M, Gozzi A, 15. Tusher VG, Tibshirani R, Chu G: Significance analysis of
Faggioni F, Domenici E, Wallace DM, Frandsen N, Polissi A: Global
microarrays applied to the ionizing radiation response. Proc Natl
analysis of transcription kinetics during competence development
Acad Sci USA 2001, 98:5116-5121.
in Streptococcus pneumoniae using high density DNA arrays. Mol
An alternative statistical method is described, designated significance analysis Microbiol 2000, 36:1279-1292.
of microarrays (SAM), for the analysis of microarray expression data. An attractive feature of this program is the use of an adjustable threshold allowingan investigator-selected false discovery rate. This method and the ANOVA- 33. De Saizieu A, Gardes C, Flint N, Wagner C, Kamber M, Mitchell TJ, based method in [14••] are available via the Internet.
Keck W, Amrein KE, Lange R: Microarray-based identification of a
novel Streptococcus pneumoniae regulon controlled by an
16. Lee M-L, Kuo FC, Whitmore GA, Sklar J: Importance of replication in
autoinduced peptide. J Bacteriol 2000, 182:4696-4703.
microarray gene expression studies: statistical methods and
Papers [32••,33••] used synthetic peptide pheromones to initiate quorum- evidence from repetitive cDNA hybridizations. Proc Natl Acad Sci
sensing developmental programs. The results demonstrate how microarray USA 2000, 97:9834-9839.
analysis can provide a genome-wide portrait of a temporally orchestrated, The inherent variability of microarray expression datasets is demonstrated, leading to the recommended use of at least three experimental replicates.
34. Throup JP, Koretke KK, Bryant AP, Ingraham KA, Chalker AF, Ge Y, Richmond CS, Glasner JD, Mau R, Jin H, Blattner FR: Genome-wide
Marra A, Wallis NG, Brown JR, Holmes DJ et al.: A genomic analysis
expression profiling in Escherichia coli K-12. Nucleic Acids Res
of two-component signal-transduction in Streptococcus
1999, 27:3821-3835.
pneumoniae. Mol Microbiol 2000, 35:566-576.
18. Zheng M, Wang X, Templeton LJ, Smulski DR, LaRossa RA, Storz G: DNA microarray-mediated transcriptional profiling of the
35. Wayne LG: Dynamics of submerged growth of Mycobacterium
Escherichia coli response to hydrogen peroxide. J Bacteriol 2001,
tuberculosis under aerobic and microaerophilic conditions. Am
183:4562-4570.
Rev Respir Dis 1976, 114:807-811.
19. Wei Y, Lee J-M, Richmond C, Blattner FR, Rafalski JA, LaRossa RA: 36. Sherman DR, Voskuil, M, Schnappinger D, Liao R, Harrell MI, High-density microarray-mediated gene expression profiling of
Schoolnik GK: Regulation of the Mycobacterium tuberculosis
Escherichia coli. J Bacteriol 2001, 183:545-556.
hypoxic response gene encoding α-crystallin. Proc Natl Acad Sci
USA 2001, 98:7534-7539.
20. Tao H, Bausch C, Richmond C, Blattner FR, Conway T: Functional
Microarray analysis of adaptation to hypoxia discloses the cognate genomics: expression analysis of Escherichia coli growing on
response regulator of the low oxygen regulon, thus demonstrating the minimal and rich media. J Bacteriol 1999, 181:6425-6440.
capacity of expression profiling to identify regulators of crucial physiologicalresponses.
21. Zimmer DP, Soupene E, Lee HL, Wendisch VF, Khodursky AB, Peter BJ, Bender RA, Kustu S: Nitrogen regulatory protein
Yuan Y, Crane D, Simpson RM, Zhu Y, Hickey MJ, Sherman DR, C-controlled genes of Escherichia coli: scavenging as a defense
Barry CE III: The 16-kDa α-crystallin (Acr) protein of
against nitrogen limitation. Proc Natl Acad Sci USA 2000,
Mycobacterium tuberculosis is required for growth in
97:14674-14679.
macrophages. Proc Natl Acad Sci USA 1998, 95:9578-9583.
22. Khodursky AB, Peter BJ, Cozzarelli NR, Botstein D, Brown PO, 38. Manganelli R, Voskuil MI, Schoolnik GK, Smith I: The
Yanofsky C: DNA microarray analysis of gene expression in
Mycobacterium tuberculosis ECF sigma factor σE: role in global
response to physiological and genetic changes that affect
gene expression and survival in macrophages. Mol Microbiol
tryptophan metabolism in Escherichia coli. Proc Natl Acad Sci
2001, 41:423-437.
USA 2000, 97:12170-12175.
A proof of principle is provided by this study of a M. tuberculosis ECF 23. Barbosa TM and Levy SB: Differential expression of over
alternative σ factor, in which it was shown how the use of σ-factor mutants 60 chromosomal genes in Escherichia coli by constitutive
and expression profiling can define regulons.
expression of MarA. J Bacteriol 2000, 182:3467-3474.
39. McKinney JD, Honer zu Bentrup K, Munoz-Elias EJ, Miczak A, Chen B, 24. Fawcett P, Eichenberger P, Losick R, Youngman P: The transcription
Chan WT, Swenson D, Sacchettini JC, Jacobs WR Jr, Russell DG: profile of early to middle sporulation in Bacillus subtilis. Proc Natl
Persistence of Mycobacterium tuberculosis in macrophages and
Acad Sci USA 2000, 97:8063-8068.
mice requires the glyoxylate shunt enzyme isocitrate lyase. Nature
2000, 406:735-738.
25. Laub MT, McAdams HH, Feldblyum T, Fraser CM, Shapiro L: Global
analysis of the genetic network controlling a bacterial cell cycle.
40. Rozwarski DA, Grant GA, Barton DH, Jacobs WR Jr, Sacchettini JC: Science 2000, 290:2144-2148.
Modification of the NADH of the isoniazid target (InhA) from
Mycobacterium tuberculosis. Science 1998, 279:98-102.
26. Mekalonos JJ: Environmental signals controlling the expression of
virulence determinants in bacteria. J Bacteriol 1992, 174:1-7.
41. Mdluli K, Slayden RA, Zhu Y, Ramaswamy S, Pan X, Mead D, Crane DD, Musser JM, Barry CE III: 1998 Inhibition of a
Paustian ML, May BJ, Kapur V: Pasteurella multocida gene
Mycobacterium tuberculosis beta-ketoacyl ACP synthase by
expression in response to iron limitation. Infect Immun 2001,
isoniazid. Science 1998, 280: 1607-1610.
69:4109-4115.
This is the first published genome-scale microarray expression study of 42. Wilson M, DeRisi J, Kristensen HH, Imboden P, Rane S, Brown PO, iron deprivation. The results reveal unanticipated complexity, suggesting a Schoolnik GK: Exploring drug-induced alterations in gene
system-wide adaptation that involves interdependent metabolic pathways.
expression in Mycobacterium tuberculosis by microarray
hybridization.
28. Gorden J, Small PL: Acid resistance in enteric bacteria. Infect
Proc Natl Acad Sci USA 1999, 96:12833-12838.
Immun 1993, 61:364-367.
43. Salama N, Guillemin K, McDaniel TK, Sherlock G, Tompkins L, 29. Arnold CA, McElhanon J, Lee A, Leonhart R, Siegele DA: Global
Falkow S: A whole-genome microarray reveals genetic diversity
analysis of Escherichia coli gene expression during the
among Helicobacter pylori strains. Proc Natl Acad Sci USA 2000,
acetate-induced acid tolerance response. J Bacteriol 2001,
97:14668-14673.
A comprehensive comparative genome database of multiple H. pylori strains Adaptation to growth in acetate by the acid-resistant enterohemorrhagic and the use of a clustering algorithm identified genes of unknown function E. coli biotype is shown to involve multiple homeostatic adjustments, some likely to have been co-inherited with the cag pathogenicity island and, thus, not previously identified by other methods of investigation.
possibly encoding new virulence determinants.
30. Perna NT, Plunkett G III, Burland V, Mau B, Glasner JD, Rose DJ, 44. Wassarman KM, Repoila F, Rosenow C, Storz G, Gottesman S: Mayhew GF, Evans PS, Gregor J, Kirkpatrick HA et al.: Genome
Identification of novel small RNAs using comparative genomic
sequence of enterohemorrhagic Escherichia coli O157:H7. Nature
microarrays. Genes Dev 2001, 15:1637-1651.
2001, 409:529-533.
45. Behr MA, Wilson MA, Gill WP, Salamon H, Schoolnik GK, Rane S, 31. Ang S, Lee C-Z, Peck K, Sindici M, Matrubutham U, Gleeson MA, Small PM: Comparative genomics of BCG vaccines by
Wang JT: Acid-induced gene expression in Helicobacter pylori:
whole-genome DNA microarray. Science 1999, 284:1520-1523.
Host–microbe interactions: bacteria
46. Kato-Maeda M, Rhee JT, Gingeras TR, Salamon H, Drenkow J, 51. Cohen P, Bouaboula M, Bellis M, Baron V, Jbilo O, Poinot-Chazel C, Smittipat N, Small PM: Comparing genomes within the species
Galiegue S, Hadibi EH, Casellas P: Monitoring cellular responses
Mycobacterium tuberculosis. Genome Res 2001, 11:547-554.
to Listeria monocytogenes with oligonucleotide arrays. J Biol
This paper and [45] demonstrate how microarray-based comparative genomics Chem 2000, 275:11181-11190.
can measure variability within natural populations of the same species and A high-density Affymetrix human array containing 6800 genes showed that between closely related species, and lead to the identification of deletions that Listeria monocytogenes induces not only a generalized innate immune may explain differences in infectivity, virulence and immunogenicity.
response in a macrophage cell line, but also pathogen-specific genes coding Alm RA, Ling L-S, Moir DT, King BL, Brown ED, Doig PC, Smith DR, for actin-binding protein and tubulin, reflecting recruitment and reorganization Noonan B, Guild BC, deJonge BL et al.: Genomic-sequence
of cytoskeletal proteins by Listeria during the process of invasion.
comparison of two unrelated isolates of the human gastric
52. Ichikawa JK, Norris A, Bangera MG, Geiss GK, van ‘t Wout AB, pathogen Helicobacter pylori. Nature 1999, 397:176-180.
Bumgarner RE, Lory S: Interaction of Pseudomonas aeruginosa
48. Israel DA, Salama N, Arnold CN, Moss SF, Ando T, Wirth HP, with epithelial cells: identification of differentially regulated genes
Tham KT, Camorlinga M, Blaser MJ, Falkow S, Peek RM Jr: by expression microarray analysis of human cDNAs. Proc Natl
Helicobacter pylori strain-specific differences in genetic content,
Acad Sci USA 2000, 97:9659-9664.
identified by microarray, influence host inflammatory responses.
Pseudomonas aeruginosa cells attached to a type II pneumocyte cell line J Clin Invest 2000, 107:611-620.
induced a variety of genes, including the gene for interferon regulatory factor 1(IRF-1), a previously described positive regulator of inducible nitric oxide 49. Rosenberger CM, Scott MG, Gold MR, Hancock RE, Finlay BB: synthase (iNOS). This result provides a possible explanation for the presence Salmonella typhimurium infection and lipopolysaccharide
of nitric oxide in the infected sputa of cystic fibrosis patients.
stimulation induce similar changes in macrophage gene
expression. J Immun 2000, 164:5894-5904.
53. Belcher CE, Drenkow J, Kehoe B, Gingeras TR, McNamara N, A murine array containing 588 genes and expressed sequence tags (ESTs) Lemjabbar H, Basbaum C, Relman DA: The transcriptional response
identified macrophage genes induced by Salmonella typhimurium and of respiratory epithelial cells to Bordetella pertussis reveal host
purified lipopolysaccharide. Both induced previously reported components defensive and pathogen counter-defensive strategies. Proc Natl
Acad Sci USA 2000, 97:13847-13852.
50. Detweiler C, Cunanan DB, Falkow S: Host microarray analysis
This paper reports a comprehensive, high-density, oligonucleotide, reveals a role for the Salmonella response regulator phoP in
Affymetrix human array study of the interaction of Bordetella pertussis with human macrophage cell death. Proc Natl Acad Sci USA 2001,
a transformed bronchial epithelial cell line. The study showed that, in addition 98:5850-5855.
to a generalized NFκB-dependent, pro-inflammatory response, the following An array containing 22 571 human genes and ESTs was used to study the genes were indicated: genes specifically activated by pertussis toxin; genes macrophage response to S. typhimurium and to an isogenic phoP mutant.
coding for respiratory tract mucins; and genes that specifically encode anti- PhoP was found to be required for the induction of macrophage genes apoptosis factors. The resulting expression profile provides a transcriptional associated with programmed cell death.
portrait that is congruent with the histopathology of whooping cough in man.

Source: http://genome.cbs.dtu.dk/staff/dave/genomics_course/FunctGenomicsPathBact_CurrOpinMicro2002.pdf