- Email: [email protected]
Analysis of host responses to microbial infection using gene expression profiling
246
Analysis of host responses to microbial infection using gene expression profiling Martin F Kagnoff* and Lars Eckmann† Gene expression profiling offers new opportunities for understanding host-cell responses to microbial pathogens and their products. Current strategies involve either first identifying mRNAs that differ in their expression status under different experimental conditions and later defining the identity of the respective genes (for example, differential display or serial analysis of gene expression), or alternatively assessing changes in the expression of already defined genes (for example, cDNA or oligonucleotide microarrays). Early studies indicate the power of gene expression profiling for providing new insights into groups of genes whose expression is altered during the course of host–microbe interactions, and for the discovery of cellular genes that were not previously recognized to be regulated by infection. Addresses University of California, San Diego, Laboratory of Mucosal Immunology, 9500 Gilman Drive, La Jolla, California 92093-0623, USA *e-mail: [email protected] †e-mail: [email protected] Correspondence: Martin F Kagnoff Current Opinion in Microbiology 2001, 4:246–250 1369-5274/01/$ — see front matter © 2001 Elsevier Science Ltd. All rights reserved. Abbreviations LPS κB NFκ SAGE
lipopolysaccharide nuclear factor κB serial analysis of gene expression
Introduction New techniques for the large-scale analysis of gene expression in host cells have opened an enormous window of opportunity for studying cellular responses to defined perturbations. In particular, global gene expression analysis is a powerful tool for studying the interactions between host cells and microbial pathogens or their products that occur at the host–microbe interface, including those at mucosal epithelial surfaces (which are key initial sites of the host–pathogen interface). Analysis of those events will provide new insights into mechanisms of microbial virulence and key host-defense mechanisms. It is both important and exciting to note that the experimental and practical applications of gene expression analysis to defining host responses to a wide range of environmental microbes are still in their infancy. Thus far, proof of principle has been established for the application of global gene expression analysis to studies of host–microbe interactions using several different technical approaches. Essentially, all of the initial studies have employed in vitro models using well-characterized invasive bacterial pathogens and cultured cell lines of epithelial cells,
monocytes/macrophages, or dendritic cells. Earlier reports on changes in gene expression in such experimental systems mostly employed differential screening or differential display techniques, although these studies were often limited to the evaluation of a small number of genes (less than 10). More recently, oligonucleotide or cDNA microarray technologies have become available for studies of host–microbe interactions, and this advance has increased the number of genes that can be studied in parallel. Nonetheless, published studies using microarrays for global gene expression analysis have focused, thus far, on a relatively limited number of genes (usually 1000–10,000), whereas the human genome is estimated to contain ~30,000 genes. These early, more focused, approaches reflect the objective of initial studies to establish the feasibility of expression profiling in various model systems of microbial infection, coupled with the realities of high cost, the limited availability, at that time, of oligonucleotide or cDNA arrays containing larger numbers of genes, and the enormous complexities of data analysis and verification. At this time, available data on gene expression profiling in host–microbe interactions are limited, especially when one considers the large potential number of different relevant microbial pathogens and their products that can be studied, and the range of host cell types that serve as potential targets for these microbes. Nonetheless, some early paradigms are emerging and the practical applications and ultimate power of gene expression profiling to gain insights into key aspects of host microbial pathogenesis are starting to become apparent. To gain an appreciation of the potential importance of the newer discovery-based approaches for studying microbial pathogenesis, it is worthwhile, for example, to consider the new insights on the functions of specific bacterial virulence genes that are likely to be gained from studies of host cell responses to wild-type bacteria, compared to relevant isogenic mutants. Although the focus of global gene expression studies thus far has been on employing systems for analysis of cellular responses in vitro, mostly using cell lines thought to represent the target cells of various microbial pathogens, important data will ultimately emerge from the application of these newer techniques to in vivo models. Interpretation of such data will not be simple, however, if one considers that, within a single organ site such as a mucosal surface, a microbial pathogen or its products may differentially alter gene expression in one or multiple cell types, each providing unique as well as overlapping responses. It will be important for in vivo studies to integrate studies on gene expression profiling with other approaches that facilitate single-cell resolution of gene expression events.
Analysis of host responses to microbial infection using gene expression profiling Kagnoff and Eckmann
Approaches to gene expression profiling Several specific methods are currently available for gene expression profiling. These utilize two general types of approaches to identify differentially expressed genes. One approach uses the strategy of first identifying mRNAs that differ in their expression status between cells, and only later defining the identity of the respective genes. This open-ended approach has the advantage of being able to identify expressed genes that may not have been cloned or are only partially sequenced. Early techniques that fall under this category are ‘differential screening’ and ‘differential display’, and more recent approaches include serial analysis of gene expression (SAGE) [1,2•]. Such openended approaches to the analysis of cellular gene expression are labor-intensive and, because of this, tend to limit the investigator to studying only a few samples and experimental conditions. In the second approach, which is currently more widely used, the identity of the genes is defined first, and their expression status is subsequently evaluated using cDNA or oligonucleotide microarrays. The breadth of information obtained from these latter methods is predetermined by the selection of known sequences included in the microarray analysis. Although each approach has clear advantages and disadvantages, both approaches have been used to discover differentially expressed genes in a wide range of biological systems. Differential screening and differential display are mentioned briefly in the following section as models of open-ended systems for studying gene expression that have been used to address host–microbe interactions in the past. In contrast, SAGE is highlighted as an approach that will likely be more broadly used in the future. cDNA and oligonucleotide microarray analysis is highlighted as the current approach that is being used most extensively in studies designed to broadly assess changes in cellular gene expression in host cells in response to microbial infection.
Differential screening and differential display In its simplest form, differential screening probes a relevant cDNA library with sets of labeled cDNAs obtained under various experimental conditions (e.g. control cells compared to cells infected with bacteria). Clones that show different labeling intensities after hybridization with the different labeled cDNAs are then selected for further analysis and are subsequently identified by sequencing. To increase the frequency of differentially expressed genes in the initial cDNA library, subtractive hybridization, which depletes genes that are not differentially expressed and enriches for genes that are differentially expressed, is commonly used. Differential screening combined with subtractive hybridization was one of the first methods reported for the identification of differentially expressed genes, but is now less commonly applied, for a number of reasons such as technical complexity, a bias towards abundantly expressed genes, and the need for large amounts of starting RNA. A more recent alternative to differential screening is the technique of differential display, and its many variations,
247
which include differential display reverse transcriptionPCR (DDRT-PCR), arbitrarily-primed PCR (AP-PCR), RNA fingerprinting, and total gene expression analysis (TOGA) [3–6]. These methods use sets of random primers (with or without anchors) to amplify by PCR unknown genes from the different sets of cDNAs. The mixtures of PCR products are then size-fractionated in parallel on a sequencing gel or by high-presence liquid chromatography, and band intensities are compared visually between different cDNA samples. Bands that differ in intensity are purified, re-amplified, and subsequently cloned and sequenced. Variations between the methods mostly pertain to the choice of primers and the amplification conditions. These methods are relatively straightforward technically and a number of commercial kits for carrying out these analyses are available. Furthermore, a relatively limited set of primers, in theory, allows one to detect nearly all genes expressed in a given cell. However, in practice, a number of limitations are often encountered — in particular, a high degree of false positive results. Although differential screening and differential display techniques are useful in identifying differentially expressed mRNAs, substantial subsequent efforts are often required to identify the respective genes. Moreover, the outcome of such studies is not predictable. A successfully identified gene may be a previously well-characterized gene, or may be unknown with no indication of its functions. The former finding is easier to pursue, as it presents a starting point for functional questions. In the optimal scenario, a gene is identified whose functions are partially known but whose involvement in the system under study is not known. On the other hand, if a gene of unknown function (a ‘new’ gene) is identified through differential screening or differential display techniques, follow-up studies can be difficult because experimental tools such as antibodies are often lacking, and useful functional studies may not be immediately apparent. As an example, early studies using differential display techniques revealed previously unidentified macrophage genes whose expression was upregulated or downregulated after the uptake of Listeria monocytogenes or mutants of L. monocytogenes into a mouse macrophage cell line [7], although further analysis was limited because many of the affected genes had no homology to known genes. In contrast, in another study in which differential display methods were used to identify differentially expressed genes in intestinal epithelial cells from germ-free mice and mice recolonised with bacteria, the upregulated ‘new’ genes turned out to be well-known genes of the cryptdin family, when cDNAs were sequenced and compared to known data bank sequences [8].
Serial analysis of gene expression The ultimate goal of mRNA expression analysis is a complete knowledge of every mRNA molecule in a given cell — the ‘transcriptome’ of that cell. In theory, this is most reliably achieved by cloning and sequencing every mRNA transcipt in a given cell. However, that amount of sequencing
248
Techniques
is currently impractical and could only be accomplished by a few research centers. Consequently, alternative approaches have been designed that reduce sequencing efforts by several 100-fold through the use of short sequence tags, instead of complete cDNAs. For example, serial analysis of gene expression (SAGE) uses restriction enzymes to obtain random 10–12 nucleotide sequence tags from the 3′ mRNA end, which are concatemerized and sequenced [1,2•]. This tag length is sufficiently long to uniquely represent all genes of the human genome, assuming that different mRNAs do not share homologies in that particular region. This is likely to be true for the vast majority of genes. Analysis of a sufficient number of random sequence tags (e.g. 50,000–100,000 tags) covers essentially all mRNAs in a given cell, and allows determination of the frequency of specific mRNAs. Parallel analysis of samples obtained under different experimental conditions will reveal differentially expressed genes. Although SAGE is complete in its approach to mRNA analysis, its major drawback is the time and resources needed for nucleotide sequence analysis of 1–2 Mbp. At present, there are few published studies that use SAGE to evaluate changes in mRNA expression in cells in response to infection with microbes, although this approach has been applied to HIV-1-infected T cells [9] and more recently to lipopolysaccharide (LPS)-stimulated human monocytes and human monocyte-derived dendritic cells [10,11•].
Microarray analysis In contrast to gene expression analysis that uses the openended approaches described above, microarray analysis starts with genes that have already been identified (although their functions may not be known) and determines the expression status of the genes under various experimental conditions (e.g. infected versus uninfected cells, or effects of wild-type bacterial strains versus isogenic mutants). In this approach, cDNA or oligonucleotide probes are immobilized on solid supports, after which labeled complex cellular RNA is hybridized to the probe, a process termed ‘hybrid selection’. Hybrid selection can be readily adapted to the analysis of the mRNA expression status of a large number of genes, as many probes can be spotted onto a suitable solid phase (typically a membrane or glass slide) and specific mRNAs are readily detected by hybridization of labeled complex RNA to the immobilized probes. Although variants of this technique have been used for decades, it received a major technical boost with the development of chemical and robotic methods to deposit oligonucleotide and cDNA probes onto glass or membrane filters to generate high-density microarrays [12–15]. Results from microarray studies reflect the diversity and number of probes employed to generate the specific microarrays used in the study. Large-scale sequencing projects, including the human genome project, that use cDNA and genomic libraries have now provided the basis for the generation of large numbers of cDNA and oligonucleotide probes for developing microarrays. Thus, the number of genes represented on specific microarrays can range from a small number of well-known genes with
related functions (e.g. pro-inflammatory cytokines, transcription factors, apoptosis-related genes) to all genes present in the human genome. Because of their ease of use, broad applicability, and the commercial availability of well-characterized oligonucleotide and cDNA array platforms, microarray approaches are currently the most widely used discovery-based techniques employed to assess changes in cellular gene expression under different experimental conditions. Nonetheless, microarray analyses suffer from several shortcomings. These include the occurrence of false positive results, limited detection and discrimination sensitivity that can vary with the array platform used and, at present, the need for substantial financial and equipment resources to conduct the work at least for analyzing expression of very large numbers of genes. In addition, minor changes in gene expression may not be detected (depending on the parameters set for significance), or may not correlate with biological importance. Further, the presently limited computational tools and bioinformatics resources available for data analyses can render extensive microarray analysis and data management difficult [16]. Microarray analysis is starting to be applied as the current state-of-the-art technique of choice to ascertain specific pieces of information (e.g. which cytokine genes are upregulated in a cell following infection) and to aid broader discovery-based gene expression approaches in microbial pathogenesis (e.g. in determining what is the totality of altered gene expression in an infected cell). Earlier studies used cDNA arrays to explore altered gene expression in response to several viral infections. For example, mRNA responses of primary human fibroblasts to infection with cytomegalovirus (CMV) were characterized by using microarrays to screen the expression of ~6,600 genes [17]. Other studies of viral interactions with host cells have focused on HIV-1 infection of CD4+ T cells [18]. Several published studies have used cDNA arrays to examine host cellular responses to bacterial pathogens, particularly with a focus on epithelial cells, which are often the first host cells to encounter microbial pathogens. One of the first studies in this regard examined changes in gene expression in human intestinal epithelial cells infected with the invasive enteric pathogen, Salmonella enterica subsp. enterica serovar Dublin (Salmonella dublin) [19••]. cDNA arrays were used to define the expression of ~4,300 genes in human colon epithelial cell lines infected with S. dublin and revealed the upregulated expression of several cytokines and chemokines, kinases important for intracellular signaling, transcription factors, and HLA molecules [19••]. A representative example of the results obtained in this study is shown in Figure 1. Importantly, results with cDNA arrays for selected genes were confirmed and evaluated further, using reverse transcription PCR, Northern blot analysis as a ‘gold standard’ for mRNA expression, and various protein assays for verification that changes in mRNA expression were paralleled by changes
Analysis of host responses to microbial infection using gene expression profiling Kagnoff and Eckmann
in protein production [19••]. As observed after CMV infection of fibroblasts, and as reported in earlier studies using microarray approaches to study cellular responses to stimulation with cytokines like γ- or α-interferon, S. dublin infection induced significant changes in mRNA expression for a relatively small fraction (~5–10%) of all genes tested. It is important to note, however, that even studies that focused on limited arrays of 4,000–10,000 genes have revealed a substantial number of genes whose mRNA expression was not previously known to increase after infection. Moreover, in the case of studies with Salmonella, a number of the genes upregulated by infection were nuclear factor κB (NFκB) target genes. Perhaps equally or more important in terms of new discovery, several of the genes upregulated after infection were genes not previously known to be targets of the transcription factor NFκB [19••]. cDNA arrays have also been applied to study the interplay between Pseudomonas aeruginosa and an alveolar epithelial cell line [20], and to studies of the interaction between Bordetella pertussis and a human bronchial epithelial cell line [21•]. As with S. dublin infection of intestinal epithelial cells, B. pertussis infection of bronchial epithelial cells rapidly altered the expression of cytokine genes and a number of NFκB-regulated target genes [21•]. cDNA and oligonucleotide arrays have also been used in early studies to identify differentially expressed genes in monocytes/macrophages after infection with microbial pathogens. For example, one study examined changes in gene expression in a human promyelocytic cell line in response to infection with L. monocytogenes [22•], whereas another report explored genes upregulated in a mouse macrophage cell line infected with Salmonella typhimurium [23••]. Of interest, the latter studies revealed several genes not previously known to be involved in the host response to that pathogen and further showed that stimulation of cells with purified S. typhimurium LPS induced the expression of many of the same genes as infection with the virulent bacteria, demonstrating the importance of LPS itself in the host response to intact bacteria [23••].
Evaluation of different approaches to gene expression profiling The choice of method selected to characterize differential gene expression in any given study is determined by several parameters, including the question being asked and the reliability, technical complexity and resources required for the various approaches. In general, there is a direct relationship between the amount of useful data generated and the resources needed to conduct the work. For example, kits for conducting differential display methods are quite affordable and easy to use, but are not likely to yield more than a few useful genes for further analysis. On the other hand, microarrays that encompass most or all human genes are likely to reveal many interesting genes in any given system, yet the purchase of arrays for such work has been beyond the resources available to many academic laboratories.
249
Figure 1
Analysis of differential gene expression in Salmonella-infected human intestinal epithelial cells. Human HT-29 intestinal epithelial cells were infected for one hour with Salmonella dublin at a multiplicity of infection of 100. The human cells were washed and further incubated with gentamicin for three hours. Uninfected cells were used as a control. RNA was extracted, reverse-transcribed in the presence of [32P]-labeled nucleotides, and subjected to hybrid selection using a membrane-bound cDNA microarray containing 268 known human cytokine and receptor genes (Atlas Human Cytokine/Receptor Array, CLONTECH Laboratories, Palo Alto, California, USA). After washing, filters were exposed to X-ray film overnight. Four genes are highlighted by boxes: two that increase after infection (MIP-2α at position b7, and IL-8 at position j3); and two whose mRNA expression is not affected by infection (IFNγ receptor at position n5, and c-Met at position j18).
Differential screening and differential display techniques were most attractive at a time when only a small fraction of human genes was known and microarrays were not available. However, these techniques are likely to lose importance in the future because of the increasing availability of complete genome sequence information and the improving affordability of microarrays. There is little rationale for identifying mRNA expression differences for unidentified genes if the expression status of all human genes can be determined in an affordable and reliable manner. Therefore, future studies are likely to focus on the use of microarrays for determining the gene expression responses of intestinal epithelial cells and other cells to infection with microbial pathogens [24] or, in some cases, initial SAGE analysis followed by the use of customized microarrays designed on the basis of subsets of genes previously identified by SAGE [25].
Conclusions As more and more laboratories employ expression profiling methods, it is important to note that false positive results are relatively commonplace in the use of cDNA microarrays. Thus, regardless of the technical approach used for microarray analysis, confirmation of results requires careful follow-up analysis using independent methods. This can be done using quantitative reverse transcription PCR, which is most easily accomplished by real-time PCR but must be optimized for each gene and may not detect unknown splice variants, and by Northern blot analysis, which is currently the ‘gold standard’ for defining the expression of known genes and can detect splice variants but is less sensitive than reverse transcription PCR. False negative results also occur when these techniques are used, because of
250
Techniques
limited assay sensitivity for rare messages that may, nonetheless, have substantial functional significance. Moreover, analyses of changes in gene expression in response to microbial infection or microbial products are only the beginning in our quest to fully understand complex issues in host–microbe pathogenesis. More complete understanding will ultimately require not only quantitative and qualitative knowledge of gene expression, but a complete analysis of changes in cellular functions, including protein production, post-transcriptional events and posttranslational protein modifications, and knowledge of protein–protein interactions that occur within cells after their interaction with infecting microbes and microbial products. The technologies used to approach many of these issues are rapidly developing, and lead not only to the possibility and realistic hope that a significant increase in the understanding of complex host–microbe interactions will ultimately be achieved, but also to novel approaches for identifying and treating microbe-induced human disease.
human activated and mature dendritic cells through serial analysis of gene expression. Blood 2000, 96:2206-2214. 11. Suzuki T, Hashimoto S, Toyoda N, Nagai S, Yamasaki N, Dong HY, • Sakai J, Yamashita T, Nukiwa T: Matsushima comprehensive gene expression profile of LPS-stimulated human monocytes by SAGE. Blood 2000, 96:2584-2591. This paper demonstrates the use of SAGE for mRNA profiling in monocytes in response to stimulation with a bacterial cell wall product. 12. Schena M, Shalon D, Davis RW, Brown PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995, 270:467-470. 13. Duggan DJ, Bittner M, Chen Y, Meltzer P, Trent JM: Expression profiling using cDNA microarrays. Nat Genet 1999, 21:10-14. 14. Lipshutz RJ, Fodor SP, Gingeras TR, Lockhart DJ: High density synthetic oligonucleotide arrays. Nat Genet 1999, 21:20-24. 15. Iyer VR, Eisen MB, Ross DT, Schuler G, Moore T, Lee JF, Trent JM, Staudt LM, Hudson JJ, Boguski MS et al.: The transcriptional program in the response of human fibroblasts to serum. Science 1999, 283:83-87. 16. Arfin SM, Long AD, Ito ET, Tolleri L, Riehle MM, Paegle ES, Hatfield GW: Global gene expression profiling in Escherichia coli K12. The effects of integration host factor. J Biol Chem 2000, 275:29672-29684. 17.
Acknowledgements The authors’ work described in this paper was supported by National Institutes of Health grants DK35108 and DK58960 and a grant from the Crohn’s and Colitis Foundation of America.
References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as:
• of special interest •• of outstanding interest 1.
Velculescu VE, Zhang L, Vogelstein B, Kinzler KW: Serial analysis of gene expression. Science 1995, 270:484-487.
2. Velculescu VE, Vogelstein B, Kinzler KW: Analysing uncharted • transcriptomes with SAGE. Trends Genet 2000, 16:423-425. This is a useful review of the use of SAGE as a powerful tool for mRNA expression profiling that utilizes an ‘open-ended’ approach. 3
Gaede KI, Mamat U, Schlaak M, Muller-Quernheim J: Analysis of differentially regulated mRNAs in peripheral blood monocytes of berylliosis patients after in vitro stimulation. J Mol Med 2000, 78:293-299.
4.
Navarro JM, Jorcano JL: The use of arbitrarily primed polymerase chain reaction in cancer research. Electrophoresis 1999, 20:283-290.
5.
Benedetti C, Fabbri M, Sitia R, Cabibbo A: Aspects of gene regulation during the UPR in human cells. Biochem Biophys Res Commun 2000, 278:530-536.
6
Sutcliffe JG, Foye PE, Erlander MG, Hilbush BS, Bodzin LJ, Durham JT, Hasel KW: TOGA: an automated parsing technology for analyzing expression of nearly all genes. Proc Natl Acad Sci USA 2000, 97:1976-1981.
7.
Schwan WR, Kugler S, Schuller S, Kopecko DJ, Goebel W: Detection and characterization by differential PCR of host eukaryotic cell genes differentially transcribed following uptake of intracellular bacteria. Infect Immun 1996, 64:91-99.
8.
Ogawa H, Fukushima K, Sasaki I, Matsuno S: Identification of genes involved in mucosal defense and inflammation associated with normal enteric bacteria. Am J Physiol Gastrointest Liver Physiol 2000, 279:G492-G499.
9.
Ryo A, Suzuki Y, Ichiyama K, Wakatsuki T, Kondoh N, Hada A, Yamamoto M, Yamamoto N: Serial analysis of gene expression in HIV-1-infected T cell lines. FEBS Lett 1999, 462:182-186.
10. Hashimoto SI, Suzuki T, Nagai S, Yamashita T, Toyoda N, Matsushima K: Identification of genes specifically expressed in
Zhu H, Cong JP, Mamtora G, Gingeras T, Shenk T: Cellular gene expression altered by human cytomegalovirus: global monitoring with oligonucleotide arrays. Proc Natl Acad Sci USA 1998, 95:14470-14475.
18. Geiss GK, Bumgarner RE, An MC, Agy MB, van Hammersmark E, Carter VS, Upchurch D, Mullins JI, Katze MG: Large-scale monitoring of host cell gene expression during HIV-1 infection using cDNA microarrays. Virology 2000, 266:8-16. 19 ••
Eckmann L, Smith JR, Housley MP, Dwinell MB, Kagnoff MF: Analysis by high density cDNA arrays of altered gene expression in human intestinal epithelial cells in response to infection with the invasive enteric bacteria Salmonella. J Biol Chem 2000, 275:14084-14094. This comprehensive study uses different sets of microarrays to characterize mRNA expression profiles in intestinal epithelial cells in response to infection with an invasive bacteria, and follows up the findings with detailed validation techniques on the mRNA and protein levels. 20. Ichikawa JK, Norris A, Bangera MG, Geiss GK, van ‘t Wout AB, Bumgarner RE, Lory S: Interaction of Pseudomonas aeruginosa with epithelial cells: identification of differentially regulated genes by expression microarray analysis of human cDNAs. Proc Natl Acad Sci USA 2000, 97:9659-9664. 21 •
Belcher CE, Drenkow J, Kehoe B, Gingeras TR, McNamara N, Lemjabbar H, Basbaum C, Relman DA: The transcriptional response of respiratory epithelial cells to Bordetella pertussis reveal host defensive and pathogen counter-defensive strategies. Proc Natl Acad Sci USA 2000, 97:13847-13852. This paper uses microarrays to study mRNA gene expression in a bronchial epithelial cell line after infection, and demonstrates the importance of NFκB as a key regulator of epithelial cell inflammatory mediators. 22. Cohen P, Bouaboula M, Bellis M, Baron V, Jbilo O, Poinot-Chazel C, • Galiegue S, Hadibi EH, Casellas P: Monitoring cellular responses to Listeria monocytogenes with oligonucleotide arrays. J Biol Chem 2000, 275:1181-11190. This study analyzes the reproducibility of microarrays in assessing macrophage mRNA responses to infection with intracellular bacteria. 23. Rosenberger CM, Scott MG, Gold MR, Hancock RE, Finlay BB: •• Salmonella typhimurium infection and lipopolysaccharide stimulation induce similar changes in macrophage gene expression. J Immunol 2000, 164:5894-904. This study compares mouse macrophage mRNA responses between infection with live bacteria and stimulation with bacterial LPS. 24. Cummings CA, Relman DA: Using DNA microarrays to study hostmicrobe interactions. Emerg Infect Dis 2000, 6:513-525. 25. Nacht M, Ferguson AT, Zhang W, Petroziello JM, Cook BP, Gao YH, Maguire S, Riley D, Coppola G, Landes GM et al.: Combining serial analysis of gene expression and array technologies to identify genes differentially expressed in breast cancer. Cancer Res 1999, 59:5464-5470.
Analysis of host responses to microbial infection using gene expression profiling Martin F Kagnoff* and Lars Eckmann† Gene expression profiling offers new opportunities for understanding host-cell responses to microbial pathogens and their products. Current strategies involve either first identifying mRNAs that differ in their expression status under different experimental conditions and later defining the identity of the respective genes (for example, differential display or serial analysis of gene expression), or alternatively assessing changes in the expression of already defined genes (for example, cDNA or oligonucleotide microarrays). Early studies indicate the power of gene expression profiling for providing new insights into groups of genes whose expression is altered during the course of host–microbe interactions, and for the discovery of cellular genes that were not previously recognized to be regulated by infection. Addresses University of California, San Diego, Laboratory of Mucosal Immunology, 9500 Gilman Drive, La Jolla, California 92093-0623, USA *e-mail: [email protected] †e-mail: [email protected] Correspondence: Martin F Kagnoff Current Opinion in Microbiology 2001, 4:246–250 1369-5274/01/$ — see front matter © 2001 Elsevier Science Ltd. All rights reserved. Abbreviations LPS κB NFκ SAGE
lipopolysaccharide nuclear factor κB serial analysis of gene expression
Introduction New techniques for the large-scale analysis of gene expression in host cells have opened an enormous window of opportunity for studying cellular responses to defined perturbations. In particular, global gene expression analysis is a powerful tool for studying the interactions between host cells and microbial pathogens or their products that occur at the host–microbe interface, including those at mucosal epithelial surfaces (which are key initial sites of the host–pathogen interface). Analysis of those events will provide new insights into mechanisms of microbial virulence and key host-defense mechanisms. It is both important and exciting to note that the experimental and practical applications of gene expression analysis to defining host responses to a wide range of environmental microbes are still in their infancy. Thus far, proof of principle has been established for the application of global gene expression analysis to studies of host–microbe interactions using several different technical approaches. Essentially, all of the initial studies have employed in vitro models using well-characterized invasive bacterial pathogens and cultured cell lines of epithelial cells,
monocytes/macrophages, or dendritic cells. Earlier reports on changes in gene expression in such experimental systems mostly employed differential screening or differential display techniques, although these studies were often limited to the evaluation of a small number of genes (less than 10). More recently, oligonucleotide or cDNA microarray technologies have become available for studies of host–microbe interactions, and this advance has increased the number of genes that can be studied in parallel. Nonetheless, published studies using microarrays for global gene expression analysis have focused, thus far, on a relatively limited number of genes (usually 1000–10,000), whereas the human genome is estimated to contain ~30,000 genes. These early, more focused, approaches reflect the objective of initial studies to establish the feasibility of expression profiling in various model systems of microbial infection, coupled with the realities of high cost, the limited availability, at that time, of oligonucleotide or cDNA arrays containing larger numbers of genes, and the enormous complexities of data analysis and verification. At this time, available data on gene expression profiling in host–microbe interactions are limited, especially when one considers the large potential number of different relevant microbial pathogens and their products that can be studied, and the range of host cell types that serve as potential targets for these microbes. Nonetheless, some early paradigms are emerging and the practical applications and ultimate power of gene expression profiling to gain insights into key aspects of host microbial pathogenesis are starting to become apparent. To gain an appreciation of the potential importance of the newer discovery-based approaches for studying microbial pathogenesis, it is worthwhile, for example, to consider the new insights on the functions of specific bacterial virulence genes that are likely to be gained from studies of host cell responses to wild-type bacteria, compared to relevant isogenic mutants. Although the focus of global gene expression studies thus far has been on employing systems for analysis of cellular responses in vitro, mostly using cell lines thought to represent the target cells of various microbial pathogens, important data will ultimately emerge from the application of these newer techniques to in vivo models. Interpretation of such data will not be simple, however, if one considers that, within a single organ site such as a mucosal surface, a microbial pathogen or its products may differentially alter gene expression in one or multiple cell types, each providing unique as well as overlapping responses. It will be important for in vivo studies to integrate studies on gene expression profiling with other approaches that facilitate single-cell resolution of gene expression events.
Analysis of host responses to microbial infection using gene expression profiling Kagnoff and Eckmann
Approaches to gene expression profiling Several specific methods are currently available for gene expression profiling. These utilize two general types of approaches to identify differentially expressed genes. One approach uses the strategy of first identifying mRNAs that differ in their expression status between cells, and only later defining the identity of the respective genes. This open-ended approach has the advantage of being able to identify expressed genes that may not have been cloned or are only partially sequenced. Early techniques that fall under this category are ‘differential screening’ and ‘differential display’, and more recent approaches include serial analysis of gene expression (SAGE) [1,2•]. Such openended approaches to the analysis of cellular gene expression are labor-intensive and, because of this, tend to limit the investigator to studying only a few samples and experimental conditions. In the second approach, which is currently more widely used, the identity of the genes is defined first, and their expression status is subsequently evaluated using cDNA or oligonucleotide microarrays. The breadth of information obtained from these latter methods is predetermined by the selection of known sequences included in the microarray analysis. Although each approach has clear advantages and disadvantages, both approaches have been used to discover differentially expressed genes in a wide range of biological systems. Differential screening and differential display are mentioned briefly in the following section as models of open-ended systems for studying gene expression that have been used to address host–microbe interactions in the past. In contrast, SAGE is highlighted as an approach that will likely be more broadly used in the future. cDNA and oligonucleotide microarray analysis is highlighted as the current approach that is being used most extensively in studies designed to broadly assess changes in cellular gene expression in host cells in response to microbial infection.
Differential screening and differential display In its simplest form, differential screening probes a relevant cDNA library with sets of labeled cDNAs obtained under various experimental conditions (e.g. control cells compared to cells infected with bacteria). Clones that show different labeling intensities after hybridization with the different labeled cDNAs are then selected for further analysis and are subsequently identified by sequencing. To increase the frequency of differentially expressed genes in the initial cDNA library, subtractive hybridization, which depletes genes that are not differentially expressed and enriches for genes that are differentially expressed, is commonly used. Differential screening combined with subtractive hybridization was one of the first methods reported for the identification of differentially expressed genes, but is now less commonly applied, for a number of reasons such as technical complexity, a bias towards abundantly expressed genes, and the need for large amounts of starting RNA. A more recent alternative to differential screening is the technique of differential display, and its many variations,
247
which include differential display reverse transcriptionPCR (DDRT-PCR), arbitrarily-primed PCR (AP-PCR), RNA fingerprinting, and total gene expression analysis (TOGA) [3–6]. These methods use sets of random primers (with or without anchors) to amplify by PCR unknown genes from the different sets of cDNAs. The mixtures of PCR products are then size-fractionated in parallel on a sequencing gel or by high-presence liquid chromatography, and band intensities are compared visually between different cDNA samples. Bands that differ in intensity are purified, re-amplified, and subsequently cloned and sequenced. Variations between the methods mostly pertain to the choice of primers and the amplification conditions. These methods are relatively straightforward technically and a number of commercial kits for carrying out these analyses are available. Furthermore, a relatively limited set of primers, in theory, allows one to detect nearly all genes expressed in a given cell. However, in practice, a number of limitations are often encountered — in particular, a high degree of false positive results. Although differential screening and differential display techniques are useful in identifying differentially expressed mRNAs, substantial subsequent efforts are often required to identify the respective genes. Moreover, the outcome of such studies is not predictable. A successfully identified gene may be a previously well-characterized gene, or may be unknown with no indication of its functions. The former finding is easier to pursue, as it presents a starting point for functional questions. In the optimal scenario, a gene is identified whose functions are partially known but whose involvement in the system under study is not known. On the other hand, if a gene of unknown function (a ‘new’ gene) is identified through differential screening or differential display techniques, follow-up studies can be difficult because experimental tools such as antibodies are often lacking, and useful functional studies may not be immediately apparent. As an example, early studies using differential display techniques revealed previously unidentified macrophage genes whose expression was upregulated or downregulated after the uptake of Listeria monocytogenes or mutants of L. monocytogenes into a mouse macrophage cell line [7], although further analysis was limited because many of the affected genes had no homology to known genes. In contrast, in another study in which differential display methods were used to identify differentially expressed genes in intestinal epithelial cells from germ-free mice and mice recolonised with bacteria, the upregulated ‘new’ genes turned out to be well-known genes of the cryptdin family, when cDNAs were sequenced and compared to known data bank sequences [8].
Serial analysis of gene expression The ultimate goal of mRNA expression analysis is a complete knowledge of every mRNA molecule in a given cell — the ‘transcriptome’ of that cell. In theory, this is most reliably achieved by cloning and sequencing every mRNA transcipt in a given cell. However, that amount of sequencing
248
Techniques
is currently impractical and could only be accomplished by a few research centers. Consequently, alternative approaches have been designed that reduce sequencing efforts by several 100-fold through the use of short sequence tags, instead of complete cDNAs. For example, serial analysis of gene expression (SAGE) uses restriction enzymes to obtain random 10–12 nucleotide sequence tags from the 3′ mRNA end, which are concatemerized and sequenced [1,2•]. This tag length is sufficiently long to uniquely represent all genes of the human genome, assuming that different mRNAs do not share homologies in that particular region. This is likely to be true for the vast majority of genes. Analysis of a sufficient number of random sequence tags (e.g. 50,000–100,000 tags) covers essentially all mRNAs in a given cell, and allows determination of the frequency of specific mRNAs. Parallel analysis of samples obtained under different experimental conditions will reveal differentially expressed genes. Although SAGE is complete in its approach to mRNA analysis, its major drawback is the time and resources needed for nucleotide sequence analysis of 1–2 Mbp. At present, there are few published studies that use SAGE to evaluate changes in mRNA expression in cells in response to infection with microbes, although this approach has been applied to HIV-1-infected T cells [9] and more recently to lipopolysaccharide (LPS)-stimulated human monocytes and human monocyte-derived dendritic cells [10,11•].
Microarray analysis In contrast to gene expression analysis that uses the openended approaches described above, microarray analysis starts with genes that have already been identified (although their functions may not be known) and determines the expression status of the genes under various experimental conditions (e.g. infected versus uninfected cells, or effects of wild-type bacterial strains versus isogenic mutants). In this approach, cDNA or oligonucleotide probes are immobilized on solid supports, after which labeled complex cellular RNA is hybridized to the probe, a process termed ‘hybrid selection’. Hybrid selection can be readily adapted to the analysis of the mRNA expression status of a large number of genes, as many probes can be spotted onto a suitable solid phase (typically a membrane or glass slide) and specific mRNAs are readily detected by hybridization of labeled complex RNA to the immobilized probes. Although variants of this technique have been used for decades, it received a major technical boost with the development of chemical and robotic methods to deposit oligonucleotide and cDNA probes onto glass or membrane filters to generate high-density microarrays [12–15]. Results from microarray studies reflect the diversity and number of probes employed to generate the specific microarrays used in the study. Large-scale sequencing projects, including the human genome project, that use cDNA and genomic libraries have now provided the basis for the generation of large numbers of cDNA and oligonucleotide probes for developing microarrays. Thus, the number of genes represented on specific microarrays can range from a small number of well-known genes with
related functions (e.g. pro-inflammatory cytokines, transcription factors, apoptosis-related genes) to all genes present in the human genome. Because of their ease of use, broad applicability, and the commercial availability of well-characterized oligonucleotide and cDNA array platforms, microarray approaches are currently the most widely used discovery-based techniques employed to assess changes in cellular gene expression under different experimental conditions. Nonetheless, microarray analyses suffer from several shortcomings. These include the occurrence of false positive results, limited detection and discrimination sensitivity that can vary with the array platform used and, at present, the need for substantial financial and equipment resources to conduct the work at least for analyzing expression of very large numbers of genes. In addition, minor changes in gene expression may not be detected (depending on the parameters set for significance), or may not correlate with biological importance. Further, the presently limited computational tools and bioinformatics resources available for data analyses can render extensive microarray analysis and data management difficult [16]. Microarray analysis is starting to be applied as the current state-of-the-art technique of choice to ascertain specific pieces of information (e.g. which cytokine genes are upregulated in a cell following infection) and to aid broader discovery-based gene expression approaches in microbial pathogenesis (e.g. in determining what is the totality of altered gene expression in an infected cell). Earlier studies used cDNA arrays to explore altered gene expression in response to several viral infections. For example, mRNA responses of primary human fibroblasts to infection with cytomegalovirus (CMV) were characterized by using microarrays to screen the expression of ~6,600 genes [17]. Other studies of viral interactions with host cells have focused on HIV-1 infection of CD4+ T cells [18]. Several published studies have used cDNA arrays to examine host cellular responses to bacterial pathogens, particularly with a focus on epithelial cells, which are often the first host cells to encounter microbial pathogens. One of the first studies in this regard examined changes in gene expression in human intestinal epithelial cells infected with the invasive enteric pathogen, Salmonella enterica subsp. enterica serovar Dublin (Salmonella dublin) [19••]. cDNA arrays were used to define the expression of ~4,300 genes in human colon epithelial cell lines infected with S. dublin and revealed the upregulated expression of several cytokines and chemokines, kinases important for intracellular signaling, transcription factors, and HLA molecules [19••]. A representative example of the results obtained in this study is shown in Figure 1. Importantly, results with cDNA arrays for selected genes were confirmed and evaluated further, using reverse transcription PCR, Northern blot analysis as a ‘gold standard’ for mRNA expression, and various protein assays for verification that changes in mRNA expression were paralleled by changes
Analysis of host responses to microbial infection using gene expression profiling Kagnoff and Eckmann
in protein production [19••]. As observed after CMV infection of fibroblasts, and as reported in earlier studies using microarray approaches to study cellular responses to stimulation with cytokines like γ- or α-interferon, S. dublin infection induced significant changes in mRNA expression for a relatively small fraction (~5–10%) of all genes tested. It is important to note, however, that even studies that focused on limited arrays of 4,000–10,000 genes have revealed a substantial number of genes whose mRNA expression was not previously known to increase after infection. Moreover, in the case of studies with Salmonella, a number of the genes upregulated by infection were nuclear factor κB (NFκB) target genes. Perhaps equally or more important in terms of new discovery, several of the genes upregulated after infection were genes not previously known to be targets of the transcription factor NFκB [19••]. cDNA arrays have also been applied to study the interplay between Pseudomonas aeruginosa and an alveolar epithelial cell line [20], and to studies of the interaction between Bordetella pertussis and a human bronchial epithelial cell line [21•]. As with S. dublin infection of intestinal epithelial cells, B. pertussis infection of bronchial epithelial cells rapidly altered the expression of cytokine genes and a number of NFκB-regulated target genes [21•]. cDNA and oligonucleotide arrays have also been used in early studies to identify differentially expressed genes in monocytes/macrophages after infection with microbial pathogens. For example, one study examined changes in gene expression in a human promyelocytic cell line in response to infection with L. monocytogenes [22•], whereas another report explored genes upregulated in a mouse macrophage cell line infected with Salmonella typhimurium [23••]. Of interest, the latter studies revealed several genes not previously known to be involved in the host response to that pathogen and further showed that stimulation of cells with purified S. typhimurium LPS induced the expression of many of the same genes as infection with the virulent bacteria, demonstrating the importance of LPS itself in the host response to intact bacteria [23••].
Evaluation of different approaches to gene expression profiling The choice of method selected to characterize differential gene expression in any given study is determined by several parameters, including the question being asked and the reliability, technical complexity and resources required for the various approaches. In general, there is a direct relationship between the amount of useful data generated and the resources needed to conduct the work. For example, kits for conducting differential display methods are quite affordable and easy to use, but are not likely to yield more than a few useful genes for further analysis. On the other hand, microarrays that encompass most or all human genes are likely to reveal many interesting genes in any given system, yet the purchase of arrays for such work has been beyond the resources available to many academic laboratories.
249
Figure 1
Analysis of differential gene expression in Salmonella-infected human intestinal epithelial cells. Human HT-29 intestinal epithelial cells were infected for one hour with Salmonella dublin at a multiplicity of infection of 100. The human cells were washed and further incubated with gentamicin for three hours. Uninfected cells were used as a control. RNA was extracted, reverse-transcribed in the presence of [32P]-labeled nucleotides, and subjected to hybrid selection using a membrane-bound cDNA microarray containing 268 known human cytokine and receptor genes (Atlas Human Cytokine/Receptor Array, CLONTECH Laboratories, Palo Alto, California, USA). After washing, filters were exposed to X-ray film overnight. Four genes are highlighted by boxes: two that increase after infection (MIP-2α at position b7, and IL-8 at position j3); and two whose mRNA expression is not affected by infection (IFNγ receptor at position n5, and c-Met at position j18).
Differential screening and differential display techniques were most attractive at a time when only a small fraction of human genes was known and microarrays were not available. However, these techniques are likely to lose importance in the future because of the increasing availability of complete genome sequence information and the improving affordability of microarrays. There is little rationale for identifying mRNA expression differences for unidentified genes if the expression status of all human genes can be determined in an affordable and reliable manner. Therefore, future studies are likely to focus on the use of microarrays for determining the gene expression responses of intestinal epithelial cells and other cells to infection with microbial pathogens [24] or, in some cases, initial SAGE analysis followed by the use of customized microarrays designed on the basis of subsets of genes previously identified by SAGE [25].
Conclusions As more and more laboratories employ expression profiling methods, it is important to note that false positive results are relatively commonplace in the use of cDNA microarrays. Thus, regardless of the technical approach used for microarray analysis, confirmation of results requires careful follow-up analysis using independent methods. This can be done using quantitative reverse transcription PCR, which is most easily accomplished by real-time PCR but must be optimized for each gene and may not detect unknown splice variants, and by Northern blot analysis, which is currently the ‘gold standard’ for defining the expression of known genes and can detect splice variants but is less sensitive than reverse transcription PCR. False negative results also occur when these techniques are used, because of
250
Techniques
limited assay sensitivity for rare messages that may, nonetheless, have substantial functional significance. Moreover, analyses of changes in gene expression in response to microbial infection or microbial products are only the beginning in our quest to fully understand complex issues in host–microbe pathogenesis. More complete understanding will ultimately require not only quantitative and qualitative knowledge of gene expression, but a complete analysis of changes in cellular functions, including protein production, post-transcriptional events and posttranslational protein modifications, and knowledge of protein–protein interactions that occur within cells after their interaction with infecting microbes and microbial products. The technologies used to approach many of these issues are rapidly developing, and lead not only to the possibility and realistic hope that a significant increase in the understanding of complex host–microbe interactions will ultimately be achieved, but also to novel approaches for identifying and treating microbe-induced human disease.
human activated and mature dendritic cells through serial analysis of gene expression. Blood 2000, 96:2206-2214. 11. Suzuki T, Hashimoto S, Toyoda N, Nagai S, Yamasaki N, Dong HY, • Sakai J, Yamashita T, Nukiwa T: Matsushima comprehensive gene expression profile of LPS-stimulated human monocytes by SAGE. Blood 2000, 96:2584-2591. This paper demonstrates the use of SAGE for mRNA profiling in monocytes in response to stimulation with a bacterial cell wall product. 12. Schena M, Shalon D, Davis RW, Brown PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995, 270:467-470. 13. Duggan DJ, Bittner M, Chen Y, Meltzer P, Trent JM: Expression profiling using cDNA microarrays. Nat Genet 1999, 21:10-14. 14. Lipshutz RJ, Fodor SP, Gingeras TR, Lockhart DJ: High density synthetic oligonucleotide arrays. Nat Genet 1999, 21:20-24. 15. Iyer VR, Eisen MB, Ross DT, Schuler G, Moore T, Lee JF, Trent JM, Staudt LM, Hudson JJ, Boguski MS et al.: The transcriptional program in the response of human fibroblasts to serum. Science 1999, 283:83-87. 16. Arfin SM, Long AD, Ito ET, Tolleri L, Riehle MM, Paegle ES, Hatfield GW: Global gene expression profiling in Escherichia coli K12. The effects of integration host factor. J Biol Chem 2000, 275:29672-29684. 17.
Acknowledgements The authors’ work described in this paper was supported by National Institutes of Health grants DK35108 and DK58960 and a grant from the Crohn’s and Colitis Foundation of America.
References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as:
• of special interest •• of outstanding interest 1.
Velculescu VE, Zhang L, Vogelstein B, Kinzler KW: Serial analysis of gene expression. Science 1995, 270:484-487.
2. Velculescu VE, Vogelstein B, Kinzler KW: Analysing uncharted • transcriptomes with SAGE. Trends Genet 2000, 16:423-425. This is a useful review of the use of SAGE as a powerful tool for mRNA expression profiling that utilizes an ‘open-ended’ approach. 3
Gaede KI, Mamat U, Schlaak M, Muller-Quernheim J: Analysis of differentially regulated mRNAs in peripheral blood monocytes of berylliosis patients after in vitro stimulation. J Mol Med 2000, 78:293-299.
4.
Navarro JM, Jorcano JL: The use of arbitrarily primed polymerase chain reaction in cancer research. Electrophoresis 1999, 20:283-290.
5.
Benedetti C, Fabbri M, Sitia R, Cabibbo A: Aspects of gene regulation during the UPR in human cells. Biochem Biophys Res Commun 2000, 278:530-536.
6
Sutcliffe JG, Foye PE, Erlander MG, Hilbush BS, Bodzin LJ, Durham JT, Hasel KW: TOGA: an automated parsing technology for analyzing expression of nearly all genes. Proc Natl Acad Sci USA 2000, 97:1976-1981.
7.
Schwan WR, Kugler S, Schuller S, Kopecko DJ, Goebel W: Detection and characterization by differential PCR of host eukaryotic cell genes differentially transcribed following uptake of intracellular bacteria. Infect Immun 1996, 64:91-99.
8.
Ogawa H, Fukushima K, Sasaki I, Matsuno S: Identification of genes involved in mucosal defense and inflammation associated with normal enteric bacteria. Am J Physiol Gastrointest Liver Physiol 2000, 279:G492-G499.
9.
Ryo A, Suzuki Y, Ichiyama K, Wakatsuki T, Kondoh N, Hada A, Yamamoto M, Yamamoto N: Serial analysis of gene expression in HIV-1-infected T cell lines. FEBS Lett 1999, 462:182-186.
10. Hashimoto SI, Suzuki T, Nagai S, Yamashita T, Toyoda N, Matsushima K: Identification of genes specifically expressed in
Zhu H, Cong JP, Mamtora G, Gingeras T, Shenk T: Cellular gene expression altered by human cytomegalovirus: global monitoring with oligonucleotide arrays. Proc Natl Acad Sci USA 1998, 95:14470-14475.
18. Geiss GK, Bumgarner RE, An MC, Agy MB, van Hammersmark E, Carter VS, Upchurch D, Mullins JI, Katze MG: Large-scale monitoring of host cell gene expression during HIV-1 infection using cDNA microarrays. Virology 2000, 266:8-16. 19 ••
Eckmann L, Smith JR, Housley MP, Dwinell MB, Kagnoff MF: Analysis by high density cDNA arrays of altered gene expression in human intestinal epithelial cells in response to infection with the invasive enteric bacteria Salmonella. J Biol Chem 2000, 275:14084-14094. This comprehensive study uses different sets of microarrays to characterize mRNA expression profiles in intestinal epithelial cells in response to infection with an invasive bacteria, and follows up the findings with detailed validation techniques on the mRNA and protein levels. 20. Ichikawa JK, Norris A, Bangera MG, Geiss GK, van ‘t Wout AB, Bumgarner RE, Lory S: Interaction of Pseudomonas aeruginosa with epithelial cells: identification of differentially regulated genes by expression microarray analysis of human cDNAs. Proc Natl Acad Sci USA 2000, 97:9659-9664. 21 •
Belcher CE, Drenkow J, Kehoe B, Gingeras TR, McNamara N, Lemjabbar H, Basbaum C, Relman DA: The transcriptional response of respiratory epithelial cells to Bordetella pertussis reveal host defensive and pathogen counter-defensive strategies. Proc Natl Acad Sci USA 2000, 97:13847-13852. This paper uses microarrays to study mRNA gene expression in a bronchial epithelial cell line after infection, and demonstrates the importance of NFκB as a key regulator of epithelial cell inflammatory mediators. 22. Cohen P, Bouaboula M, Bellis M, Baron V, Jbilo O, Poinot-Chazel C, • Galiegue S, Hadibi EH, Casellas P: Monitoring cellular responses to Listeria monocytogenes with oligonucleotide arrays. J Biol Chem 2000, 275:1181-11190. This study analyzes the reproducibility of microarrays in assessing macrophage mRNA responses to infection with intracellular bacteria. 23. Rosenberger CM, Scott MG, Gold MR, Hancock RE, Finlay BB: •• Salmonella typhimurium infection and lipopolysaccharide stimulation induce similar changes in macrophage gene expression. J Immunol 2000, 164:5894-904. This study compares mouse macrophage mRNA responses between infection with live bacteria and stimulation with bacterial LPS. 24. Cummings CA, Relman DA: Using DNA microarrays to study hostmicrobe interactions. Emerg Infect Dis 2000, 6:513-525. 25. Nacht M, Ferguson AT, Zhang W, Petroziello JM, Cook BP, Gao YH, Maguire S, Riley D, Coppola G, Landes GM et al.: Combining serial analysis of gene expression and array technologies to identify genes differentially expressed in breast cancer. Cancer Res 1999, 59:5464-5470.