0889–857X/02 $15.00 .00
GENETICS
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY Jerry A. Molitor, MD, PhD, Jane H. Buckner, MD, and Gerald T. Nepom, MD, PhD
Transcript array analysis is a technique that has emerged from the explosive growth in the field of genomics. This technique examines the expression of thousands of genes simultaneously. Transcript arrays have been used to identify the biologic programs that determine the yeast cell cycle, yeast sporulation, and response of human cells to stimulation with serum or cytokines. These approaches are now being applied to human disease. Transcript array analyses are being used to clarify the diagnosis and prognosis of malignancies and to understand the underlying pathogenesis of complex human disorders such as the rheumatic diseases. In this review, we outline the use of transcript arrays to simultaneously assess gene activation of hundreds or thousands of genes and their potential use in understanding and managing rheumatic disorders. We focus on the use of transcript arrays to confirm and refine disease diagnoses and on the opportunity that this technology provides to generate new hypotheses regarding pathophysiologic findings in rheumatic diseases. These approaches promise to lead to improvements in developing rational approaches to therapy as well as to improvements in diagnosis and prognosis, and they may allow the profiling of patients with respect to their likely response to a variety of therapies. This work was supported in part by grant AR37296 from the National Institutes of Health and by the Elmer and Mary Louise Rasmuson Center for Rheumatic Disease at Virginia Mason Research Center.
From the Section of Rheumatology (JAM, JHB), Virginia Mason Medical Center (JHB, GTN), and Department of Immunology, University of Washington School of Medicine (GTN), Seattle, Washington
RHEUMATIC DISEASE CLINICS OF NORTH AMERICA VOLUME 28 • NUMBER 1 • FEBRUARY 2002
151
152
MOLITOR et al
ARRAY TECHNOLOGY The fundamental tools of gene expression analysis reviewed here are cDNA and oligonucleotide arrays. Oligonucleotides or cDNAs are bound to a solid surface; the reverse-transcribed sample is then hybridized to the bound DNA. Early arrays were produced by spotting cDNAs onto nylon membranes. Such arrays typically contain several hundred separate cDNA spots from genes of interest. In this method, mRNA is isolated from the cells of interest and a closely matched control sample for comparison. The mRNA is then reverse-transcribed, with radioactive or fluorescent label incorporated into the cDNA. Hybridization of the reverse-transcribed mRNA samples to the array is then performed. When fluorescently labeled nucleotides of distinct emission wavelengths are incorporated into the reverse-transcribed cDNA of control and sample mRNA, hybridization is performed simultaneously on the array. This allows a comparison of the relative gene expression in the sample of interest and the control based on fluorescence intensities. Microarrays use these techniques but are capable of simultaneously allowing hybridization to thousands of genes. Each of the cDNA species of interest is spotted onto derivatized glass slides, producing a so-called ‘‘chip.’’ The technology involved in the production and use of such microarrays, first described by Brown’s laboratory at Stanford University,62 is beyond the scope of this review but has been extensively reviewed elsewhere.12, 14, 17, 20, 26, 34, 39 Hybridization of the control cDNA is performed simultaneously on the chip, and fluorescence profiles are detected by a scanning confocal microscope. Because of the large number of genes present on each chip, genes may be grouped or ‘‘clustered’’ by their level of expression in the sample of interest relative to the control. These groups can then suggest how genes may be related in biologic pathways. Oligonucleotide arrays are constructed with direct combinatorial synthesis of oligonucleotides onto a solid substrate. Oligonucleotides as small as 25 base pairs in length can be used. As described previously, fluorescently labeled cDNA samples are hybridized to the chip and detected by scanning confocal microscopy. Nonspecific hybridization is controlled by the comparison of binding between the oligonucleotides of interest and closely related oligonucleotides containing deliberately mismatched residues. Comparison of control samples with experimental samples is usually performed externally in this method, with control samples hybridized to a separate chip. More detailed descriptions of this technology and validation of its use have been reported elsewhere.41, 43, 69 CLUSTERING ALGORITHMS The simultaneous analysis of thousands of genes has presented challenges that have been addressed in several ways. The development of databases from which array data can later be analyzed, or ‘‘mined,’’
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
153
has been reviewed previously.9, 29, 67 Mathematic algorithms have also been developed to group, or ‘‘cluster,’’ data based on similarities in the data set. Several prior reviews have covered this topic.17, 66, 70 Clustering has been performed using hierarchical27 and nonhierarchical clustering algorithms. Hierarchical clustering places data sets together by grouping the closest pairs of points and replacing these two with the single point that represents the average, followed by successive iterations of this procedure until there is a hierarchy of nested subsets. Nonhierarchical cluster methods include deterministic annealing,7 K-means testing,74 selforganizing maps,72 a ‘‘fuzzy logic’’ approach,81 and a method of analysis called relevance networking.15 These techniques continue to evolve; a complete discussion of them is beyond the scope of this review, but their application to the analysis of array data provides insights into lineage or functional relations among unknown genes related by expression pattern. USE OF ARRAYS IN MODEL SYSTEMS Yeast Several studies have exemplified the utility of microarray profiling using studies in yeast. Yeast studies are important validators of the technique, because the yeast genome has been fully sequenced and the physiologic functions of many yeast genes are well understood. Although we do not comprehensively review this area, we discuss here several studies that demonstrate how microarrays have expanded the level of understanding of gene regulation in this system. In a comprehensive analysis of the yeast response to environmental changes,30 yeast stress responses to a variety of stressors such as heat shock and amino acid starvation were analyzed. A notable feature of the expression array data in these experiments was that a number of genes were coordinately expressed, and a large set of genes were coordinately repressed in response to diverse stimuli. Of 6200 genes arrayed, approximately 600 genes were repressed, and 300 were induced by most stimuli. Two distinct profiles were evident. A cluster of growthrelated genes, including genes implicated in RNA metabolism, translational initiation and elongation, tRNA synthesis and processing, nucleotide biosynthesis, and other metabolic processes were all coordinately downregulated in response to stress. Promoter analysis of the genes involved in this downregulated cluster revealed two conserved motifs in upstream elements of the genes, both of which were novel. A second cluster was distinguished by a slight delay in the decline of transcript levels and consisted almost entirely of genes encoding ribosomal proteins. Of the 300 upregulated genes, 60% were previously uncharacterized. The functions of known genes included carbohydrate metabolism, detoxification of reactive oxygen species, cellular redox reactions, cell wall modification, protein folding and degradation, DNA damage repair,
154
MOLITOR et al
and fatty acid metabolism. These gene responses were largely one-way; that is, they result in response to transition of the cells from a normal growth state to a state of increased stress and not the other way around. Transition from one stressful environment to another, however, resulted in the additive expression of genes involved in an environmental stress response. An encyclopedic analysis of genes regulated during sporulation in yeast has demonstrated the differential expression of approximately 1000 genes.22 Clustering identified seven temporally based expression profiles. The promoters of the genes in each of these clusters were analyzed, and many of the genes in each cluster were shown to have shared promoter elements. For example, genes induced early after transfer to sporulation medium were likely to contain an upstream repression sequence–1 (URS1) transcription factor promoter motif. By contrast, genes induced several hours later after transfer to sporulation medium were more likely to contain a midsporulation element (MSE) promoter element. The use of microarray analysis combined with previous knowledge of yeast transcription factors and their binding sites allowed the identification of an entire set of coordinately regulated genes. Human Fibroblast Cell Cycle Regulation The temporal program of transcriptional regulation in fibroblasts after the addition of fetal bovine serum (FBS) has been studied using cDNA expression arrays.36 Primary cultured foreskin fibroblasts were deprived of serum for 48 hours and then stimulated with 10% FBS. The mRNA levels of 8613 genes were measured at intervals ranging from 15 minutes to 24 hours after stimulation. Clustering of the expression data revealed 10 distinct patterns of expression. Among the first genes to be activated were genes known to be important in signal transduction, including c-FOS, JUNB, and the mitogen-activated protein kinase phosphatase-1, all of which were induced within 15 minutes after serum stimulation. Conversely, genes whose products inhibit progression of cell cycle, including p27 Kip1, p57 Kip2, and p18, were all downregulated by serum stimulation, with the maximal downregulation occurring between 6 and 12 hours. This analysis also revealed upregulation of a number of genes involved in inflammation, clotting and clot dissolution, chemotaxis, angiogenesis, and re-epithelialization. These include the chemokines interleukin (IL)-8, macrophage inflammatory protein-2␣ (MIP–2␣) and monocyte chemoattractant protein–1 (MCP–1), cyclooxygenase-2 (COX-2), intercellular adhesion molecule (ICAM)-1, IL-1, and IL-6 as well as factor III, endothelin-1, plasminogen activator inhibitor type 1, and vascular endothelial growth factor. By contrast, genes encoding proteins involved in the biosynthesis of cholesterol were sharply downregulated 4 to 6 hours after serum stimulation, which was thought possibly to be secondary to the presence of cholesterol within low-density lipoprotein (LDL) in serum. This important study demonstrated that serum normally en-
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
155
countered by cells in vivo in the context of a wound serves not only as a mitogen but as a proinflammatory, hemostatic, and angiogenic stimulus. Serum also promotes tissue remodeling and re-epithelialization through complex upregulation of a variety of genes. This study demonstrates the need for caution in the interpretation of in vitro studies that use culture medium supplemented with FBS, but clearly demonstrates the utility of the expression array approach in elucidating multiple complex effects of a simple stimulus. Cell cycle–regulated transcripts have also been studied in human foreskin fibroblasts.21 In this protocol, cells were synchronized in the late G1 phase, and samples were collected every 2 hours over a 24-hour period. Transcripts were identified on high-density oligonucleotide arrays containing 6800 genes. A total of 731 transcripts, including 344 previously uncharacterized transcripts, were identified as having cell cycle–regulated expression patterns. DNA replication genes were induced in late G1 and S phases. Genes promoting cytoskeletal reorganization, mitosis, cell cycle control, and muscular contraction were all upregulated during the G2 phase, and genes that regulate chromosome segregation, cell-to-cell adhesion, and mitosis and cell control were upregulated during the M phase. There was also enrichment of genes encoding proteins causing apoptosis within the G2 phase. The effects of ultraviolet light and methyl methane sulfonate, both of which cause DNA damage and cell cycle arrest, were examined with respect to the mRNA transcripts. A total of 42 transcripts were induced in response to ultraviolet light– or methyl methane sulfonate–induced damage. There was overlap between these genes and genes found in S phase upregulation but not for other phases within the cell cycle. Flow cytometry analysis was performed to demonstrate that the cell cycle arrest caused by these agents did not occur predominantly or exclusively in the S phase. MICROARRAY ANALYSIS OF IMMUNE CELLS Lymphocytes, monocytes, macrophages, and dendritic cells are key effectors and regulators of the immune system. Each of these cell types has been extensively studied ex vivo, although it is clear that their functions are extensively intertwined in the organism. Array analysis in immunology has been summarized recently in a comprehensive review.70 We focus here on selected topics that are likely to be of significant importance in the understanding of autoimmune disease. T Lymphocytes A pioneering study demonstrating transcript analysis in T and B lymphocytes5 demonstrated important differences in the activation of normal T cells by phorbol ester, phytohemagglutinin, and IL-2. A second group has examined the effects of T-cell activation with the staphylococ-
156
MOLITOR et al
cal superantigen enterotoxin B.75 After 8 hours of stimulation, 280 of 6319 genes were differentially expressed. At 48 hours after activation, however, only 51 genes differed in their level of expression. Genes with significantly decreased expression included the IL-4 receptor; IL-7␣ chain; IL-6 receptor; decay accelerating factor; leukocyte-7 transmembrane receptor; signal transduction protein MyD116; and Kruppel-like transcription factors LKLF, EZF, and NBKLF. In addition, there were dramatic decreases in the FOS/JUN family mRNA species and for the stress–signalling kinase (SEK1) MAP kinase. Results of this study and several other early genomic studies of T lymphocytes are discussed in a recent review.47 CD4 T Helper 1 versus T Helper 2 Cell Analysis A fundamental paradigm that has furthered our understanding of T-lymphocyte function and disease pathophysiologic findings is the concept of T helper (Th) 1 versus Th2 cell determinism (reviewed by Abbas et al1). Many chronic inflammatory diseases are mediated by a Th1 type of CD4 T cell as are antibacterial immune responses, whereas allergies and immune defense against helminths and other parasites seem to be mediated by Th2 type cells. Th1 cells produce interferon–␥ (IFN–␥) and are implicated in the pathogenesis of an inflammatory pannus in rheumatoid arthritis (reviewed by Miossec and van den Berg51), whereas Th2 cells produce IL-5 and IL-10. A recent study has generated data from highly purified Th1 and Th2 cells for comparative analysis by oligonucleotide arrays.60 Purified differentiated CD4 T cells were prepared from cord blood leukocytes, and the mRNA from each of five such differentiated samples was hybridized to oligonucleotide arrays. Interferon–␥; IL-12 receptor 2 (the IL-12 receptor  chain); oncostatin M; IL-18 receptor; macrophage inflammatory protein-1; the adhesion molecule; p-cadherin; and the transcription factors ets-1, nuclear factor (NF)-IL6, and c-myb were among the genes significantly upregulated in the Th1 cell subset. By contrast, IL-10 receptor, the adhesion molecule integrin ?, and the transcription factor GATA-3 were significantly downregulated in Th1 cells. IL-12 is known to be an important cofactor in the differentiation of Th1 cells, and allows their priming for maximum IFN–␥ synthesis.76 Interleukin-12 was incubated with Th1 cells, and the IL-12–treated Th1 cell mRNA was arrayed against untreated Th1 cell mRNA. As expected, there was a marked increase in interferon-␥ message. The IL-12 receptor 2 and IL-18 receptor increased approximately threefold further, and the NFIL-6, interferon regulatory factor (IRF), and NFIL-6 transcription factor mRNAs were all significantly further increased. In addition to these genes, the leukocyte adhesion gene FUT7, whose product is important in accomplishing fucosylation of selectin ligand adhesion molecules on T cells,45 was upregulated after IL-12 treatment. FUT7 mRNA transcripts were significantly upregulated in T cells purified from the synovial fluid of patients newly diagnosed with rheumatoid arthritis.
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
157
Figure 1. T cell polarization to TH1 or TH2 pathways. Cord-blood derived human CD4 T cells were studied for cytokine secretion one week following in vitro culture in the presence of cytokines. A high ratio of interferon/IL-5 following IL-12 treatment indicates TH1 lineage commitment; a reversal of this ratio following IL-4 treatment indicates TH2 lineage commitment.
We have used a similar in vitro culture system in which naı¨ve cord blood T lymphocytes are differentiated to a Th1 or Th2 lineage for microarray analysis of gene transcripts, which characterize the latter commitment events associated with these distinct phenotypes. After 7 days of culture on autologous adherent cells treated with exogenous IL-4 or IL-12, T cells were recovered, washed, and recultured with autologous dendritic cells, without additional exogenous cytokines. During this secondary culture, the T cells complete their commitment and maturation toward polarized Th1 or Th2 cells, although no exogenous cytokine stimuli are added. Figure 1 illustrates the rapid induction of polarized cytokines in these cultures, which is sustained at least through day 21. Cultures that were initially exposed to IL-12 produce large amounts of IFN–␥ but not IL-5; cultures initially exposed to IL-4 produce large amounts of IL-5 but not IFN–␥. mRNA from the Th1 and Th2 samples was labeled with either Cy3 or Cy5, respectively, to generate a fluorescent probe and was used to probe the GEM microarray (Genome Systems, Cambridge, UK). The GEM microarray used consisted of a collection of sequence-verified amplified cDNAs derived from the UniGemV NCBI database; it includes thousands of clones representing approximately 4100 known human genes as well as 3000 expressed sequence tags (ESTs). The cDNAs (500–5000 base pairs in length) are arrayed on glass chips, and data were analyzed using the Genome Systems GemTools 2.3 system in which the Cy3 or Cy5 signal strength, fluorescent area, and ratio of expression between the dyes were calibrated and compared. Twenty-two genes that were preferentially expressed in the Th2 cells at a threefold or greater ratio relative to the Th1 cells were identified, including 8 that were expressed at a fourfold or higher ratio. There were over 100 additional genes expressed at a level of twofold or greater from a total of over 7000 transcripts. An example of a portion of these arrays is shown in Figure
158
MOLITOR et al
Figure 2. Example of transcript array analysis in studies of T cell lineage commitment. Human cord blood-derived CD4 T cells were cultured in the presence of cytokines to support differentiation into TH1 (using IL-12) or TH2 (using IL-4) pathways for 7 days. One week after removal of cytokines, cDNA was prepared and labeled either with Cy-3 (TH1 cultures) or Cy-5 (TH2 cultures), and hybridized to genomic arrays. A small portion of the array is shown, encompassing an 8 12 matrix representing 96 genes. Several cDNA probes in the array, such as C3 and F2, preferentially hybridize to TH1-derived transcripts, while others such as F12 hybridize to TH2 transcripts; positive hybridization is represented as fluorescent (bright) spots on the grid.
Table 1. EXAMPLES OF GENES PREFERENTIALLY EXPRESSED IN T HELPER2–BIASED HUMAN T-CELL CULTURES, IDENTIFIED BY cDNA ARRAY ANALYSIS Differential Expression
Gene Name
4.2 4.0 4.0 3.7 3.7 3.6 3.6 3.5 3.3 3.3 3.2 3.1 2.7 2.7
IL-4 receptor EST:U90916 MAL EST:D87465 CCR7 EST:AFO38178 EST:N29083 GILZ EST:L47738 L-selectin Phosphodiesterase 7A EST:N41047 IL-7 receptor IL-10 receptor
*mRNA for the analysis was derived from human cord blood T cells that had acquired a T helper (Th)2 phenotype in culture, compared with similar Th1 mRNA. Known transcripts identified included expected (i.e., interleukin (IL)-4 receptor, IL-10 receptor) and unexpected (i.e., MAL, GILZ) genes. Among the latter, functional clues are suggested by the array findings; for instance, GILZ has previously been implicated as a transcription factor associated with resistance to apoptosis, and its expression in Th2 cells may correspond to the strong differences between Th1 and Th2 cells in sensitivity to activation-induced cell death. Several of the unknown genes identified in the array are candidates for contributing to specific Th2 pathways or phenotypes as well. IL interleukin; EST expressed sequence tag.
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
159
2, and a list of several Th2 candidate genes of interest is shown in Table 1. Confirmatory studies were performed using mRNA from additional cord blood samples, using reverse-transcriptase polymerase chain reaction analyses for expression of the candidate Th2-expressed genes. Further confidence in the validity of the Th2 transcript array study comes from the fact that several of the genes identified such as the IL-4 receptor and IL-10 receptor are clearly expected outcomes and serve as an internal positive control. In the study by Rogge et al60 discussed previously, mRNA was collected on day 3 of the cultures, during the early lineage determination phase, and before the commitment phase. Interestingly, some of the same genes that they identified expressed early (day 3) are identical to genes we identified late (day 14); other genes differed between the two time points. B Lymphocytes An interesting use of microarrays in B lymphocytes has been the analysis of the transcriptional state during tolerance. Mice transgenic for the production of hen egg lysozyme (HEL) induce tolerance in their B cells that produce HEL-specific immunoglobulin.31 The authors used oligonucleotide expression arrays to monitor 6500 genes between tolerant and naı¨ve B-cell preparations. The expression of 20 genes was significantly increased, and that of 8 genes was decreased in the tolerant cells. In parallel, the expression patterns of naı¨ve B cells were measured in response to the HEL stimulus. This showed the activation of 37 genes and the decrease in expression of 22 genes after 1 hour of activation. After 6 hours of activation, however, an additional 300 genes increased and 200 decreased their mRNA expression, many of them consistent with the movement of the cell to the G1 phase. Direct comparison of the activation response with the tolerance response showed that almost all the genes upregulated in activation were blocked in tolerance. Of the more than 500 transcript changes caused by antigen stimulation of naı¨ve B cells within 6 hours of activation, only 16 were also induced by antigen stimulation of tolerant cells. Activation of naı¨ve B cells in the presence of FK506 suppressed only one third of the 59 transcript changes detected 1 hour after B lymphocyte activation. These array analyses clarify mechanisms and pathways that underlie the requirement for continuous immunosuppression with medicine such as FK506 and those that may be involved in tolerance, thereby indicating targets of potential utility for the resetting of tolerance with future medications. A group of investigators led by Staudt and his colleagues at the National Cancer Institute have used arrays to better characterize B-cell activation states. Using 14407 B-cell cDNA clones together with 3446 other cDNA clones of immunologic importance, they created a cDNA microarray called a lymphochip.4 With this tool, initial experiments were performed on malignant and normal B cells. Comparison of various Bcell activation states revealed a large number of genes coordinately
160
MOLITOR et al
expressed in germinal center cells that were minimally or undetectably expressed in peripheral blood cells. Most of these genes were not induced, however, during in vitro culturing of peripheral blood B cells with anti-IgM antibodies together with CD40 ligand and IL-4. These stimuli induce mitogenesis in peripheral blood B cells and activate a variety of known B-cell genes but did not recapitulate the gene activation program seen in germinal center B cells. Among the genes differentially expressed between activated peripheral blood B cells and germinal center B cells were the chemokine receptors CXCR5 and CCR7, both of which are much more highly expressed than activated peripheral B cells on germinal center cells.3 The antiapoptotic protein bcl-2 is also profoundly downregulated in germinal center B cells compared with peripheral activated B cells. These studies likely have broad implications for rheumatology given the involvement of B cells in autoimmune diseases. The presence of germinal centers in the rheumatoid synovium82 and the characterization of B-cell differentiation within these germinal centers64 underscore the likelihood that genomic approaches that monitor B-cell activity in synovium are of potential clinical utility. This approach may also be important in the diagnosis of Sjo¨gren’s syndrome and in the prognostication of those rheumatoid arthritis and Sjo¨gren’s syndrome patients at risk of Bcell lymphoma. Human Myeloid Cells The transcriptional program of immature and mature dendritic cells has been compared using cDNA microarrays containing 4110 known genes.25 These authors used CD14-selected peripheral blood mononuclear cells to generate mature dendritic cells through in vitro incubation with granulocyte macrophage colony-stimulating factor (GM-CSF) and IL-4. Mature dendritic cells were obtained by further culturing for 3 days in the presence of IL- 6, tumor necrosis factor (TNF)-␣, IL-1, and prostaglandin E2. Comparisons of these two cell types demonstrated an increase of 291 transcripts by at least twofold and a decrease in 78 transcripts to less than one half of their prior level in the mature dendritic cells compared with the immature dendritic cells. Genes with increased expression in mature dendritic cells included the GM-CSF␣ chain, the IL-4 receptor, the IL-7 receptor-␣ and IL-15 receptor-␣ chains, the insulin-like growth factor-1 receptor, IL-1␣, IL-6, and transforming growth factor-␣. There was a dramatic increase in expression of the chemokine TARC as well as in that of the chemokine receptor CCR7 in mature dendritic cells and smaller increases in MCP-1 and MCP-2. By contrast, the chemokine receptors CCR1 and CCR5 were increased in immature dendritic cells compared with mature dendritic cells. The IL1 receptor type 2 was also increased in the immature dendritic cell population as was the colony stimulating factor (CSF–1) receptor. Immature dendritic cells also showed increased levels of expression of adhe-
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
161
sion proteins, including CD11B and galactin-3, as well increased levels of genes implicated in antigen processing and presentation. A recent study presented by O’Shea and colleagues at the National Institutes of Health has examined total RNA extracted from human peripheral mononuclear cells that were stimulated with a variety of distinct cytokines, including IL-2, IL-4, IL-7, IL-15, IL-12, and IL-18.19 A 6800-gene array was used. A total of 456 genes were found to be significantly regulated by one or more of these cytokines after 7 hours of stimulation. Hierarchic clustering algorithms were used to examine the data. Genes stimulated by IL-2 and IL-15 were relatively similar and differed from genes stimulated by IL-4. IL-18 had a distinct genome expression profile, and IL-12 plus IL-18 had a synergistic effect on the stimulation of gene transcription, producing significant induction of 68 genes not induced by either of these cytokines singly.19 In summary, transcript array analysis of isolated T-cell, B-cell, and myeloid populations has been fruitful for the detailed characterization of specific genomic patterns and individual genes that correlate with specific activation and maturation states. These studies have identified genomic profiles for cellular responses that can be used as benchmarks in the more complex setting involving clinical samples that are heterogeneous with respect to lineage and stimuli. Cellular Transcriptional Response to Infection An interesting and important aspect of our understanding of the immune response is the issue of how individual cell types respond to an infection. The study of the global transcriptional response of the cell during infection is of substantial importance to our understanding of the possible linkage of autoimmune diseases to antecedent infection. Several studies using transcript arrays in this area have been reviewed recently.46 In one particularly interesting study, primary human foreskin fibroblasts were infected with human cytomegalovirus, and the mRNA from these cells was compared with that of uninfected cells. The transcripts were probed with oligonucleotide arrays containing greater than 6600 genes.85 RNA preparations compared at 40 minutes after either mock or viral infection demonstrated the level of 27 mRNAs changed by a factor of 3 or more. The number of differentially expressed mRNA species increased to 93 at 8 hours and to 364 at 24 hours after infection. Interestingly, several genes were identified whose altered regulation by cytomegalovirus is of potential importance in the induction of autoimmune disease. The nonclassic class 1 molecule, human leukocyte antigenE, which allows natural killer cell recognition of the target cell without killing,11, 13 was upregulated approximately sixfold. Upregulation of this gene may allow cytomegalovirus-infected cells to escape destruction and potentially allow persistent viral infection. Cytosolic phospholipase A2 mRNA increased 12-fold, COX-2 mRNA increased sevenfold, and lipo-
162
MOLITOR et al
cortin-1 mRNA decreased ninefold at 24 hours after infection. The concerted altered regulation of these genes would likely result in a marked increase in production of the inflammatory mediator prostaglandin–E2, which should, in turn, be controllable with COX-2 inhibitors. Of further interest to rheumatologists was the demonstration that the Ro/SSA 52-kd protein mRNA was induced by factor of 12 at 24 hours after infection in response to cytomegalovirus infection. Although these authors have not shown protein overexpression in their system, this degree of mRNA overexpression does raise the distinct possibility that either correctly or alternately spliced Ro/SSA proteins would be much more abundantly expressed in cytomegalovirus-infected cells. The infected cells are likely to be provoking an immune response, and overexpressed Ro/SSA may be more likely to be seen as an autoantigen under these conditions. Chlamydia-infected epithelial (HeLa) cells have also been examined using cDNA microarrays.23 In this study, IL-8 and IL-11 were increased after infection with the serovar L2. Of note, there seemed to be little upregulation of IL-1 or TNF␣. An additional study has examined the alteration in gene expression produced by human papilloma virus (HPV) type 31 in HPV31transfected human keratinocytes compared with nontransfected control keratinocytes.18 HPV31 expression caused an increase in the expression of 178 genes between two- and threefold, and 150 genes were found to be repressed by at least twofold. Most of the genes that showed increased activation were unknown ESTs. By contrast, however, a number of interferon-inducible genes were repressed by HPV31, including interferon––inducible 11.5-kd protein, interferon-inducible protein 56, and Stat 1. In addition, IL-1 receptor type 2 was significantly downregulated. Because Stat 1 is a major regulator of interferon-responsive genes, the authors examined the levels of Stat 1 protein in the HPV31-positive cells and found that induction of Stat 1 protein in response to interferon-␣ and interferon-␥ was reduced compared with normal human keratinocytes. As interferons ␣ and  become an increasingly important therapy for various diseases, and as therapies directed at minimizing interferon–␥ based response in diseases such as rheumatoid arthritis come to the fore, understanding of these pathways should be of increasing importance. USE OF EXPRESSION ARRAYS IN HUMAN DISEASE Analysis of Human Cancers One of the great promises of array technology in clinical medicine is the possibility for it to improve diagnosis and clarify prognosis. Most human autoimmune diseases are characterized by a variety of clinical features that generally display remarkable heterogeneity in their expression between individual patients. Microarray analysis of gene expression
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
163
offers promise in allowing us to potentially improve our existing clinical pattern recognition by adding the pattern recognition of gene expression on a genomic scale. The attempt to subset complex human disease states with this type of analysis is in its infancy; it has, however, been successfully carried out in studies of malignancies. Microarray technology is of clear utility in dissecting out patterns of gene expression that further define certain types of cancers. The technology is also showing great promise in identifying specific genes that may contribute to cancer pathophysiology and in defining genes that are affected by chemotherapeutic agents as well as finding patterns of gene expressions that may be linked through clustering algorithms to chemotherapeutic susceptibility. We summarize several of these examples with a focus on how they provide guidance for the conduct of similar studies in the rheumatic and autoimmune diseases. Leukemia and Lymphoma A pioneering study in the use of expression arrays for diagnosis demonstrated that this technique could distinguish acute myeloid leukemia and acute lymphoblastic leukemia based on oligonucleotide expression arrays without simultaneous immunohistochemical analysis.32 This study analyzed 38 bone marrow samples obtained from acute leukemia patients at diagnosis and found that approximately 1100 genes were more highly correlated with one or the other disease state than would have been expected by chance. A collection of the 50 genes most closely correlated with a clear distinction between the two leukemia types was then used to assign diagnosis with an independent collection of 34 leukemia samples. This was successful in making strong predictions for 29 of these 34 samples with an accuracy of 100%. Like leukemias, non-Hodgkin’s lymphomas are clinically heterogeneous. Of these, diffuse large B-cell lymphoma (DLBCL) is the most common subtype. This subtype, however, contains additional clinical heterogeneity; only 40% of patients respond well to the current therapy. A large collaborative group has used the lymphochip4 to examine DLBCL and have found two molecularly distinct forms of this tumor.6 Three types of adult lymphoid malignancies, DLBCL, follicular lymphoma (FL), and chronic lymphocytic leukemia were studied. Arrays of mRNA from FL samples showed a pattern of gene expression virtually identical to that of the germinal center B cells, suggesting that FL arises from this stage of B-cell differentiation. The gene expression profiles of the DLBCL, however, were distinct from those of chronic lymphocytic leukemia and FL. The feature of these studies that is perhaps of greatest interest to clinicians is the demonstration that the molecularly defined DLBCL subgroups seem to define a subgroup of patients with a distinct clinical prognosis. Using Kaplan-Meyer plots of the overall survival of the DLBCL patients and subgrouping all patients based on the DLBCL profiles, 76% of germinal center B-cell–like DLBCL patients were alive
164
MOLITOR et al
after 5 years compared with only 16% of activated B-cell–like DLBCL patients. These studies have examined thousands of genes and identified patterns of gene expression that distinguish different types of leukemias and lymphomas. Expression arrays have also been used to identify a new diagnostic marker for anaplastic large cell lymphoma (ALCL).80 A cDNA expression array consisting of 588 human cDNA fragments was used to identify a gene, clusterin, specifically found in 4 ALCL cell lines but in none of 27 other tumor cell lines. Of 198 primary lymphomas, 36 were ALCL, and all 36 demonstrated clusterin expression by immunohistochemistry and Western blotting. By contrast, 160 of 162 non-ALCL lymphomas were clusterin-negative. This study demonstrates that array analysis can contribute to improved diagnosis by identifying altered expression of previously unknown or unsuspected genes that may suggest new immunohistochemical markers. Breast Cancer Several studies have now explored the use of microarrays in the analysis of human breast cancers.48, 56, 57 As might be anticipated given the histologic complexity of the tissue, sets of transcripts characteristic of extracellular matrix, inflammation, and cell proliferation have all been identified. Clinical utility is suggested by the finding of transcript profiles that were associated with estrogen receptor status, clinical stage, or tumor size.48 In this study, a gene cluster that included heat shock protein 90 was overexpressed in stage 4 tumors, which are distinguished from lower stage tumors by the presence of distant metastasis. In a pioneering use of arrays to monitor response to therapy,57 transcripts have been studied from tumors that were sampled twice, first after the open surgical biopsy and then after 16-week treatment with doxorubicin, followed by resectioning of any remaining tumor at that time. Hierarchic clustering was used to group genes on the basis of similarity in their expression patterns. Although there was wide variation in the patterns of gene expression between different tumors, 15 of 20 before and after doxorubicin-treated samples clustered closely together despite the doxorubicin treatment and an interval of 16 weeks between the sampling. Of note, in 3 tumors, the specimens obtained after chemotherapy clustered together with 3 normal breast samples. Retrospective analysis of the clinical data demonstrated that these 3 tumors were doxorubicin ‘‘responders.’’ As has been seen in other studies of gene expression in cancer cells, there was a ‘‘proliferation cluster,’’ containing many genes known to be expressed during cellular proliferation. Expression in this cluster varied widely among the tumor samples, although it was generally well correlated with mitotic index. In 36 of 38 samples, variation of estrogen receptor-alpha gene mRNA correlated well with estrogen receptor protein levels. Eight independent clusters of genes
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
165
seemed to reflect variation in specific cell types present within the tissue. These included an endothelial cell cluster, a stromal cell cluster, an adipose-enriched/normal breast cell cluster, a B-lymphocyte cluster, a Tlymphocyte cluster, and a macrophage cluster. Two additional clusters distinguished between basal epithelial cells and luminal epithelial cells. There was also a cluster of tumors that were characterized by high-level expression of a subset of genes, including the Erb-B2 oncogene as well as low level expression of estrogen receptor. In summary, the authors were able to identify four groups of samples that were potentially related to different molecular features of mammary epithelial biology (the estrogen receptor–positive/luminal-like, estrogen receptor–negative/basallike, estrogen receptor–negative/Erb-B2 positive, and normal breast). The ability of this and other studies to identify the probable presence of distinct cell types by clustering is of substantial importance for attempts to extend these techniques to complex tissues. Malignant Melanoma cDNA microarray analysis has also been employed in analyzing melanomas.10 Samples from a number of patients were analyzed, including two pairs of specimens derived from the same patients at different time points. Transcript arrays from these samples were highly correlated with each other, despite the fact that one of these two pairs represented two different samples from the same individual that were surgically removed over a year apart. The second of these samples was a biopsy sample for which mRNA was directly prepared compared with a cell culture sample of the same tumor that was carried for three passages in vitro. These results are encouraging in that they suggest that gene expression patterns within an individual patient’s tissues may be rather remarkably conserved over time. Cluster analysis of the genes involved in melanoma samples showed that there were two major clusters, one of which showed reduced expression of a variety of adhesion molecule genes that would predict reduced mobility and reduced invasive ability such as integrin 1, integrin 3, integrin ␣1, syndecan 4, and vinculin. This transcript pattern correlated with functional parameters of reduced invasive ability and reduced vasculogenic mimicry, and there was a trend toward reduced mortality in the patients within this cluster. Other Cancers The application of microarray analysis to other human cancers is rapidly expanding. Studies on ovarian cancer50, 65, 78 prostate cancer,16, 28 cervical cancer,68 colon cancer,8 and squamous cell carcinoma of the lung79 have each demonstrated the differential expression of various genes between cancer cells and control cells. Taken together, and with the ability to identify the presence of possible contaminating cells based
166
MOLITOR et al
on their gene expression signatures (as demonstrated by Perou and his colleagues56, 57), it is clear that the use of cDNA expression microarrays is important in the analysis of a variety of tumor types.
Microarray Analysis of Tumor Susceptibility to Antineoplastic Agents Several studies have now examined the ability of clustering techniques to infer relations between chemotherapeutic sensitivity of human cancer cell lines and the individual gene expression profiles of each of those lines. Scherf et al63 focused on a 118-drug subset whose mechanisms of action were putatively understood and developed five large coherent clusters that corresponded closely to mechanisms of action. These included DNA and DNA/RNA antimetabolites, tubulin inhibitors, DNA damaging agents, topoisomerase (Top) 1 inhibitors, and Top 2 inhibitors. These authors found several examples of highly significant correlations between gene expression and drug sensitivity. One of these was a negative correlation between dihydropyrimidine dehydrogenase, the rate-limiting enzyme in uracil and thymidine catabolism, and 5fluorouracil potency. They also demonstrated a moderately high negative correlation between expression of asparagine synthetase and tumor Lasparaginase sensitivity. An alternative approach used relevance network clustering.15 The authors compared the baseline expression of 6701 genes as well as measures of susceptibility to 4991 anticancer agents. Nine of the developed networks showed associations with potential biologic relations. Several of these were between closely associated genes, and one correctly linked two intermediate filament keratin proteins known to be coexpressed and which function together but have disparate sequences. One network demonstrated a relation between the gene known as lymphocyte cytosolic protein 1 and the anticancer agent NSC624044, a thiazolidine carboxylic acid derivative. The effect of chemotherapeutic agents on tumor cells can be directly measured using array analysis before and after treatment of tumor cell lines with an agent of choice. As noted previously,57 breast cancers have been analyzed before and after treatment with doxorubicin. In a separate study,40 the MCF7 breast cancer–derived cell line was transiently treated with doxorubicin and selected for resistance to it. Both of these conditions were then compared with untreated MCF7 cells using a 5600-gene microarray. A subset of genes found to be induced by doxorubicin after transient treatment was also induced in cells resistant to doxorubicin, possibly representing genes involved in the mechanisms of drug resistance. A similar approach has been taken in analyzing the differentiation program induced by retinoic acid in acute promyelocytic leukemia cells.42
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
167
MICROARRAY ANALYSIS OF AUTOIMMUNE DISEASES Cancer is a clonal disease and is well suited to transcript array studies, where gene expression levels correlate with cancer phenotype. But what of autoimmune diseases? In most diseases, we do not know with any certainly what the cell type of interest is. In rheumatoid arthritis, for example, the debate persists over the relative importance of the B lymphocytes, different T-lymphocyte subsets, macrophage-like synoviocytes, and fibroblast-like synoviocytes. Is the cell of primary importance in scleroderma the fibroblast, endothelial cell, natural killer cell, or T cell? If it is the T cell, is it the CD4 T cell, or the CD8 T cell? Even if the cell type of importance is known, can we isolate that cell type in sufficient purity? Alternatively, do clustering methods allow us to define gene expression patterns from complex tissues containing several different cell types and then use the clustering techniques to define the likely cell type of origin of the genes of interest? Although these questions remain unanswered, arrays are being increasingly employed in the analysis of autoimmune diseases.
Rheumatoid Arthritis One of the earliest published array studies examined gene expression in primary rheumatoid chondrocytes, synoviocytes, and inflamed lower intestinal mucosal samples of patients with Crohn’s disease.35 In mixed primary rheumatoid synoviocytes and chondrocytes, there was low–level expression of c–JUN, granulocyte colony–stimulating factor (GCSF), IL–3, TNF– and the chemokines. For example, migration inhibitory factor (MIF) and RANTES. There was also expression of the matrix metalloproteinases GEL A, Strom–1, and Col–1, and of the tissue inhibitors of matrix metalloproteinases (TIMP)–1, –2, and –3. The relatively low-level expression of these genes was increased after treatment of cells with the phorbol–ester (PMA), plus either IL-1 or TNF␣ plus IL-1, although to a lesser degree with the latter treatment. Other genes upregulated by these treatments included IL-6, IL-8, human matrix metalloelastase, and the adhesion molecule vascular cell adhesion molecule-1 (VCAM–1). Rheumatoid synovial samples from different affected individuals gave relatively similar profiles as did different samples from the same individual. There were distinct differences between the inflammatory bowel disease samples, and, in general, there was less marked upregulation of inflammatory genes than in rheumatoid tissues, although IL-1 converting enzyme, the metalloelastase inhibitor TIMP-1, and macrophage migration inhibitor factor were consistently upregulated in three different patient samples examined. This study was remarkable in showing that a variety of inflammatory genes were upregulated in demonstrating a rather consistent pattern between different individuals with the same disease as well as in showing consistent upregulation over time in the same individuals. Furthermore, measure-
168
MOLITOR et al
ment of the upregulation of these genes did not require purification of individual cell types. Verweij et al77 have used the lymphochip to compare synovial tissue samples from patients with rheumatoid arthritis and patients with osteoarthritis. Using tissue and synoviocyte type B-cell lines, distinct clusters were obtained demonstrating differential expression of a number of genes between the rheumatoid and osteoarthritis samples. Another group37 has also reported on the use of a cDNA expression array for differential screening using small amounts of RNA obtained from synovial fibroblasts from rheumatoid arthritis versus osteoarthritis. One of 15 differentially expressed genes, CD82, was confirmed to be expressed predominantly in fibroblasts within the synovial lining layer by doublelabeling experiments. A recently discovered tumor suppressor, P33ING1, was also found to be downregulated in synovial fibroblasts in this study. A separate group from the United Kingdom has examined the gene expression in synovial biopsies taken from patients with early inflammatory synovitis. These patients were then followed for a year, and the expression array data were compared.2 This analysis revealed that genes including fibronectin and several ESTs were consistently more abundant in self-limited disease, and three additional cDNAs were found to be significantly more frequent in persistent disease. These are unknown genes, and their sequences are being explored as possible leads in determining the prognosis of early undifferentiated inflammatory arthritis. A distinct procedure for identifying disease-related genes has been developed by the Burmester laboratory.49 This group has made a subtractive cDNA expression library. Circulating blood monocytes from a rheumatoid arthritis patient with active disease were obtained by leukophoresis and then subtracted from the cDNA expression library obtained from monocytes from the same patient after three leukophoreses, at which time, inflammatory gene expression is markedly decreased.71 Differentially expressed genes from several additional rheumatoid arthritis patients have included ferritin, IL-6, IL-8, IL-1␣ and IL1. This group has gone on to compare cDNA expression patterns using microarrays and is currently identifying novel overexpressed genes.49 The effects of estrogen stimulation of human synoviocytes have also been examined by microarry.38 A 1000-gene microarray was used to compare gene expression from 17-estradiol–treated synoviocytes from rheumatoid arthritis patients and control individuals. Several genes of interest were differentially modulated, but of particular interest was the expression of B-cell–homing chemokine transcripts in increased amounts in rheumatoid synoviocytes in response to hormone. Patterns of changes of gene expression caused by gene therapy are another application for cDNA expression arrays. An interesting study has examined gene expression from rheumatoid arthritis synovial fibroblasts after adenovirus-based TNF␣ receptor p55 gene transfer.53 The upregulation of several genes was observed, including the inhibitor of proliferation p48 and the growth inhibitory factor metallothionein (MT)–
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
169
III and the downregulation of tumor associated proto-oncogenes notch 2 and c-myc. In a similar approach, the cyclin-dependent kinase inhibitor gene was introduced into rheumatoid arthritis synovial fibroblasts, and cDNA was prepared and analyzed by cDNA expression arrays. Prior studies have demonstrated that induction of this gene in rheumatoid synovial cells and animal models decreases the production of inflammatory mediators.73 In the array analysis, 135 genes were upregulated and 193 were downregulated by at least twofold in response to cyclin-dependent kinase inhibitor gene expression. Downregulated genes include a number of cell cycle–related genes, apoptosis-related genes, transcription factors, growth factors, and cytokines.52, 54 Ulcerative Colitis A recent study has examined mucosal gene expression from ulcerative colitis patient samples using oligonucleotide arrays.24 The authors compared eight ulcerative colitis specimens and seven nonulcerative colitis samples obtained from microscopically normal areas of surgical resection for adenocarcinoma, diverticular abscess, and diverticulosis. A total of 74 mRNAs whose expression was reproducibly increased in ulcerative colitis were found. These included a variety of matrix metalloproteinases, stromolysin-1, lymphotoxin-, IL-8, the CD20 receptor, and the CD40 ligand receptor precursor. Clusters of transcripts associated with elevated disease activity scores included IL-1, IL-6, IL-8, IL-11, GRO-2, lymphotoxin-, IL1 receptor antagonist-␣, COX-2, GM-CSF, monocyte chemotactic activating factor, and ICAM-1. A variety of antigens were also found, including CD83 and CD19 (identifying dendritic cells); the B-cell differentiation antigens CD21, CD20 receptor, CD22, CD53, and CD127; the endothelial activation markers CD62E, CD62P and EDG-1; and the early T-cell activation antigen CD69. This study demonstrated successful analysis using complex tissues, with retrospective clustering algorithms used to delineate important patterns of gene expression and assignation to specific pathophysiologic cell types. Scleroderma Several groups have recently analyzed gene expression in cells from scleroderma patients using transcript microarrays. Most of these studies have used scleroderma skin fibroblasts. Zhou et al83, 84 compared RNA from systemic sclerosis (SSC) patients’ affected and unaffected SSC skin fibroblasts as well as from peripheral blood mononuclear cells and muscle biopsies. A number of autoantigen genes were overexpressed, irrespective of whether the RNA was from involved or uninvolved skin. These included fibrillarin, centromeric protein B, and RNA polymerase II. Semiquantitative reverse-transcriptase polymerase chain reaction confirmed these changes but also showed increased levels of Top 1 in the
170
MOLITOR et al
SSC fibroblasts compared with controls, a change that was not noted on the arrays. Neither eosinophilic fasciitis nor scleromyxedema fibroblasts displayed increased expression of these genes. Osteonectin (SPARC), osteonectin, metalloproteinase-3, human leukocyte antigen-C, and RANTES also showed high expression levels in SSC fibroblasts. Analysis of the peripheral blood monocytes, B cells, and muscle biopsies from SSC patients, however, did not show increases in the autoantigens seen in the fibroblasts. There was a significant upregulation of the 70-kd U1RNP gene in muscle biopsies from SSC patients compared with biopsies from five normal controls. Other investigators demonstrated differential expression of matrix metalloproteinases and TIMPs in scleroderma fibroblasts.59 Comparison of lesional and nonlesional skin fibroblasts from six scleroderma patients showed matrix metalloproteinase-3 and TIMP-1, -2, and -3 specifically upregulated and matrix metalloproteinase-1 downregulated in lesional fibroblasts. In an additional recent study, more than 60 genes were differentially expressed between scleroderma skin fibroblasts and normal skin fibroblasts, including heat shock protein 90. The upregulation of this gene has been confirmed and is of further interest, as additional experiments showed that geldamycin, an HSP90 inhibitor, blocks TGF induction of collagen synthesis in NIH–3T3 fibroblasts. This probably occurs through blockage of SMAD signal transduction proteins, as SMAD3 and SMAD4 were inhibited in geldamycin- and TGFtreated fibroblasts (J. Korn, MD, personal communication, 2001). An additional group has studied the patterns of gene expression in unfractionated bronchoalveolar lavage cells from scleroderma patients. Seventeen scleroderma patients and 7 control patients were examined. A 4000-gene microarray was used, and hierarchic clustering was performed. A total of 372 genes were differentially expressed in scleroderma patients without lung inflammation compared with controls, including several chemokine receptors and IL-1, IL-13, IL-18, and IFN–␥ receptors as well as IL-10, IL-12, macrophage stimulating factor, vascular endothelial growth factor, insulin–like growth factor (IGF) binding protein, and the TSK tyrosine kinase. Fibroblast growth factors and other cytokines as well as intracellular signaling molecules were among the genes differentially expressed in other clusters. A total of 238 genes were identified as characteristic of bronchoalveolar lavage cells from patients demonstrating greater risk of lung fibrosis compared with the SSC patients without lung inflammation and control patient samples. These genes clustered into three groups, which included stress-induced genes, chemokines, and intracellular signaling pathway modulators. Further data examining CD8 T cells suggested abnormal expression of lymphotoxin-B, endothelin-2, fibroblast growth factors, and TGF receptor 2.44 Systemic Lupus Erythematosus A microarray approach has been used to study gene expression from the peripheral blood mononuclear cells from systemic lupus erythematosus (SLE) patients.61 Using an array containing 375 genes, peripheral
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
171
blood mononuclear cells from five SLE patients and seven healthy controls were compared. Seventeen genes were identified that were differentially expressed by at least twofold. Upregulated genes included the IL10–induced chemotactic chemokine HCC4, the FLT-3 ligand, the CCR7 chemokine receptor, the adhesion molecules integrin-␣3 and p-cadherin, and the antiapoptotic gene TOSO. Downregulated genes included TNF receptor 1 and 2, interferon-␥ receptor 1, monokine induced by interferon-␥, and TNF-related apoptosis inducing ligand receptor 1. A separate group has undertaken the difficult task of analyzing the gene expression in SLE glomerular biopsies obtained by laser capture microdissection. To obtain enough RNA to perform the cDNA expression array, sample RNA was amplified using T7-BASE linear methods. Using a 5500-gene microarray, the authors found 340 genes consistently overexpressed in SLE glomeruli and 74 consistently downregulated compared with those in control biopsies.58 Among upregulated genes were enzymes mediating tissue repair and regeneration and interferon-induced genes. Seronegative Arthritides Microarrays have been used to examine the gene expression from biopsies of affected and unaffected skin from psoriasis patients.55 Using oligonucleotide arrays, over 200 genes were found to be differentially regulated in lesional versus nonlesional skin biopsies. Several genes known to be involved in psoriasis were overexpressed, including type 1 cytokine genes, validating the analysis. Over 100 genes were identified with no previous known link to the disease, however. Furthermore, peripheral blood mononuclear cell gene expression was compared between psoriasis patients and normal individuals, and more than 60 additional inflammatory genes were identified in this manner. Importantly, expression levels of a number of these genes were normalized in response to treatment. Microarrays have also been applied to the gene expression profiles of peripheral blood mononuclear cells of spondyloarthropathy (SPA) patients.33 These authors compared peripheral blood mononuclear cell gene expression profiles between rheumatoid arthritis, psoriatic arthritis, and SPA patients. They also analyzed the gene expression profiles of synovial fluid cells from rheumatoid arthritis and SPA patients. Although the peripheral blood mononuclear cells of SPA patients showed relatively few activated genes6 compared with those of rheumatoid arthritis patients,30 these did include TNF␣, IL-8, and ICAM-1. Rheumatoid arthritis and SPA patients demonstrate significant upregulation of TNF␣ and IL-1 genes in the synovial cells. FUTURE DIRECTIONS Transcript array technology has established itself as a useful tool in examining cellular responses to stimuli in model systems and has been
172
MOLITOR et al
useful in the characterization of some malignancies. The application of this technology to complex genetic diseases, including the rheumatic diseases, is only now underway. Application of this technology to diseases that are clinically heterogenous will be challenging, particularly when the pathogenic tissue and cell types are not known. Despite these difficulties, the potential for expanding our understanding of autoimmunity with these techniques is great. Arrays will improve our ability to diagnose disease, predict disease outcomes, and identify subgroups of patients responsive to certain therapies. Further use of transcript arrays is likely to shed light on the underlying mechanisms of disease, allowing the development of new therapeutic agents. What is likely to facilitate the application of these rapidly developing technologies to the clinic? Among the most important factors are the examination of samples from well-characterized groups of patients. This requires detailed databases regarding the patients’ clinical features, laboratory evaluation, and response to medications so as to allow correlation between these clinical parameters and the expression of certain genes. Identification of clusters of genes that predict prognosis is important in terms of selection of therapy and, ultimately, in the development of single nucleotide polymorphism maps that may be used for early diagnosis of complex genetic diseases. Because microarrays represent the net gene expression pattern of all cell types present in the sample, some characterization of the cell types in a tissue is desirable. This is, of course, more difficult in complex tissue specimens. A number of the analyses done on solid tumors and other tissue samples suggest that when gene expression is robust, purification of individual cell types may not be required to obtain useful data from the microarray. Clustering techniques may prove critically important in identifying tissuespecific gene expression in these situations. Nonetheless, definition of the cell types present in the tissue should be helpful when using cluster analysis to avoid misinterpretation of the results. In many of the studies published and included in this review, the authors have made their primary data available on associated web sites. This is potentially helpful, because the comparison of studies with subtle differences in design or patient populations can be envisioned. Clinical uses of array technologies are in their infancy. Clinicians can anticipate rapid extension of transcript analysis and single nucleotide polymorphism analysis using DNA chips to routine clinical practice in the not too distant future. DNA testing for diagnosis and prognosis is likely to be added to the clinical armamentarium, and the future use of DNA testing to predict responses to specific therapies and to monitor response to therapy should have a dramatic impact on patient management and the clinical course of disease. References 1. Abbas AK, Murphy KM, Sher A: Functional diversity of helper T lymphocytes. Nature 383:787–793, 1996
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
173
2. Ali M, Veale DJ, Reece RJ, et al: Comparison between persistent and self-limiting synovitis by gene expression [abstract]. Arthritis Rheum 43(suppl):S324, 2000 3. Alizadeh A, Staudt LM: Genomic-scale gene expression profiling of normal and malignant immune cells. Curr Opin Immunol 12:219–225, 2000 4. Alizadeh A, Eisen M, Davis RE, et al: The lymphochip: A specialized cDNA microarray for the genomic-scale analysis of gene expression in normal and malignant lymphocytes. Cold Spring Harb Symp Quant Biol 64:71–78, 1999 5. Alizadeh A, Eisen M, Botstein D, et al: Probing lymphocyte biology by genomic-scale gene expression analysis. J Clin Immunol 18:373–379, 1998 6. Alizadeh AA, Eisen MB, Davis RE, et al: Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 403:503–511, 2000 7. Alon U, Barkai N, Notterman DA, et al: Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc Natl Acad Sci USA 96:6745–6750, 1999 8. Backert S, Gelos M, Kobalz U, et al: Differential gene expression in colon carcinoma cells and tissues detected with a cDNA array. Int J Cancer 82:868–874, 1999 9. Bassett DE, Jr, Eisen MB, Boguski MS: Gene expression informatics—it’s all in your mine. Nat Genet 21:51–55, 1999 10. Bittner M, Meltzer P, Chen Y, et al: Molecular classification of cutaneous malignant melanoma by gene expression profiling. Nature 406:536–540, 2000 11. Borrego F, Ulbrecht M, Weiss EH, et al: Recognition of human histocompatibility leukocyte antigen (HLA)-E complexed with HLA class I signal sequence-derived peptides by CD94/NKG2 confers protection from natural killer cell-mediated lysis. J Exp Med 187:813–818, 1998 12. Bowtell DDL: Options available—from start to finish—for obtaining expression data by microarray. Nat Genet 21:25–32, 1999 13. Braud VM, Allan DS, O’Callaghan CA, et al: HLA-E binds to natural killer cell receptors CD94/NKG2A, B and C. Nature 391:795–799, 1998 14. Brown PO, Botstein D: Exploring the new world of the genome with DNA microarrays. Nat Genet 21:33–37, 1999 15. Butte AJ, Tamayo P, Slonim D, et al: Discovering functional relationships between RNA expression and chemotherapeutic susceptibility using relevance networks. Proc Natl Acad Sci USA 97:12182–12186, 2000 16. Carlisle AJ, Prabhu VV, Elkahloun A, et al: Development of a prostate cDNA microarray and statistical gene expression analysis package. Mol Carcinog 28:12–22, 2000 17. Celis JE, Kruhoffer M, Gromova I, et al: Gene expression profiling: Monitoring transcription and translation products using DNA microarrays and proteomics. FEBS Lett 480:2–16, 2000 18. Chang YE, Laimins LA: Microarray analysis identifies interferon-inducible genes and Stat-1 as major transcriptional targets of human papillomavirus type 31. J Virol 4:4174– 4182, 2000 19. Cheng TP, Ahn H-J, Galon J, et al: Analyzing gene expression by immunoregulatory cytokines through the use of microarrays [abstract]. Arthritis Rheum 43(suppl):S85, 2000 20. Cheung VG, Morley M, Aguilar F, et al: Making and reading microarrays. Nat Genet 21:15–19, 1999 21. Cho RJ, Huang M, Campbell MJ, et al: Transcriptional regulation and function during the human cell cycle. Nat Genet 27:48–54, 2001 22. Chu S, DeRisi J, Eisen M, et al: The transcriptional program of sporulation in budding yeast. Science 282:699–705, 1998 23. Dessus-Babus S, Knight ST, Wyrick PB: Chlamydial infection of polarized HeLa cells induces PMN chemotaxis but the cytokine profile varies between disseminating and non-disseminating strains. Cell Microbiol 2:317–327, 2000 24. Dieckgraefe BK, Stenson WF, Korzenik JR, et al: Analysis of mucosal gene expression in inflammatory bowel disease by parallel oligonucleotide arrays. Physiol Genomics 4:1–11, 2000 25. Dietz AB, Bulur PA, Knutson GJ, et al: Maturation of human monocyte-derived
174
26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50.
MOLITOR et al
dendritic cells studied by microarray hybridization. Biochem Biophys Res Commun 275:731–738, 2000 Duggan DJ, Bittner M, Chen Y, et al: Expression profiling using cDNA microarrays. Nat Genet 21:10–14, 1999 Eisen MB, Spellman PT, Brown PO, et al: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 95:14863–14868, 1998 Elek J, Park KH, Narayanan R: Microarray-based expression profiling in prostate tumors. In Vivo 14:173–182, 2000 Ermolaeva O, Rastogi M, Pruitt KD, et al: Data management and analysis for gene expression arrays. Nat Genet 20:19–23, 1998 Gasch AP, Spellman PT, Kao CM, et al: Genomic expression programs in the response of yeast cells to environmental changes. Mol Biol Cell 11:4241–4257, 2000 Glynne R, Akkaraju S, Healy JI, et al: How self-tolerance and the immunosuppressive drug FK506 prevent B-cell mitogenesis. Nature 403:672–676, 2000 Golub TR, Slonim DK, Tamayo P, et al: Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring. Science 286:531–537, 1999 Gu J, Kuipers J, Maerker-Hermann E, et al: Microarray gene expression profiles in spondyloarthropathy (SPA) mirror disease-mediating process [abstract]. Arthritis Rheum 43(suppl):S396, 2000 Hegde P, Qi R, Abernathy K, et al: A concise guide to cDNA microarray analysis. Biotechniques 29:548–562, 2000 Heller RA, Schena M, Chai A, et al: Discovery and analysis of inflammatory diseaserelated genes using cDNA microarrays. Proc Natl Acad Sci USA 94:2150–2155, 1997 Iyer VR, Eisen MB, Ross DT, et al: The transcriptional program in the response of human fibroblasts to serum. Science 283:83–87, 1999 Judex M, Neumann E, Kullmann F, et al: Combination of RAP-PCR for differential display and cDNA expression array identifies P33ING1 and CD82, genes hitherto unrelated to rheumatoid arthritis (RA) [abstract]. Arthritis Rheum 43(suppl):S96, 2000 Khalkhali-Ellis Z, Seftor EA, Moore TL, et al. Microarray analysis of the response of synoviocytes to 17 -estradiol [abstract]. Arthritis Rheum 43(suppl):S96, 2000 Khan J, Saal LH, Bittner ML, et al: Expression profiling in cancer using cDNA microarrays. Electrophoresis 20:223–229, 1999 Kudoh K, Ramanna M, Ravatn R, et al: Monitoring the expression profiles of doxorubicin-induced and doxorubicin-resistant cancer cells by cDNA microarray. Cancer Res 60:4161–4166, 2000 Lipshutz RJ, Fodor SP, Gingeras TR, et al: High density synthetic oligonucleotide arrays. Nat Genet 21:20–24, 1999 Liu TX, Zhang JW, Tao J, et al: Gene expression networks underlying retinoic acidinduced differentiation of acute promyelocytic leukemia cells. Blood 96:1496–1504, 2000 Lockhart DJ, Dong H, Byrne MC, et al: Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 14:1675–1680, 1996 Luzina IG, Atamas SP, Wise R, et al: Global analysis of gene expression of unseparated and CD8 cells from bronchoalveolar lavage of patients with scleroderma lung disease [abstract]. Arthritis Rheum 43(suppl):S168, 2000 Maly P, Thall A, Petryniak B, et al: The alpha(1,3)fucosyltransferase Fuc-TVII controls leukocyte trafficking through an essential role in L-, E-, and P-selectin ligand biosynthesis. Cell 86:643–653, 1996 Manger D, Relman DA: How the host ‘‘sees’’ pathogens: Global gene expression responses to infection. Curr Opin Immunol 12:215–218, 2000 Marrack P, Mitchell T, Hildeman D, et al: Genomic-scale analysis of gene expression in resting and activated T cells. Curr Opin Immunol 12:206–209, 2000 Martin KJ, Kritzman BM, Price LM, et al: Linking gene expression patterns to therapeutic groups in breast cancer. Cancer Res 60:2232–2238, 2000 Martinez-G L, Stuhlmu¨ller B, Hernandez M, et al: Analysis of differentially expressed novel genes in activated rheumatoid arthritis monocytes [abstract]. Arthritis Rheum 43(suppl):S96, 2000 Martoglio A-M, Tom BDM, Starkey M, et al: Changes in tumorigenesis- and angiogen-
TRANSCRIPT ARRAY ANALYSIS IN RHEUMATOLOGY
51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.
175
esis-related gene transcript abundance profiles in ovarian cancer detected by tailored high density cDNA arrays. Mol Med 6:750–765, 2000 Miossec P, van den Berg W: Th1/Th2 cytokine balance in arthritis. Arthritis Rheum 40:2105–2115, 1997 Nagasaka K, Kohsaka H, Nonomura Y, et al: Effects of cyclin-dependent kinase inhibitor p16INK4a gene induction of gene expression by rheumatoid synovial fibroblasts [abstract]. Arthritis Rheum 43(suppl):S99, 2000 Neumann E, Fleck M, Judex M, et al. Alteration of proto-oncogene and cytokine expression following adenovirus-based TNF-␣R p55 gene transfer in rheumatoid synovial fibroblasts [abstract]. Arthritis Rheum 43(suppl):S100, 2000 Nonomura Y, Kohsaka H, Nagasaka K, et al: Effects of cyclin-dependent kinase inhibitor P21cip1 gene induction on gene expression by rheumatoid synovial fibroblasts [abstract]. Arthritis Rheum 43(suppl):S99, 2000 Oestreicher JL, Walters IB, Gilleaudeau P, et al: Pharmacogenomic identification of psoriasis and inflammatory related disease [abstract]. FASEB J 14:A1213, 2000 Perou CM, Jeffrey SS, van de RM, et al: Distinctive gene expression patterns in human mammary epithelial cells and breast cancers. Proc Natl Acad Sci USA 96:9212–9217, 1999 Perou CM, Sorlie T, Eisen MB, et al: Molecular portraits of human breast tumours. Nature 406:747–752, 2000 Peterson KS, Zhu J, D’Agati V, et al: Pathogenesis of SLE nephritis: An approach to its study using DNA microarrays [abstract]. Arthritis Rheum 43(suppl):S395, 2000 Power CA, Shi-wen X, Colinge J, et al: cDNA array technology reveals that specific members of the matrix metalloprotease family and their inhibitors are differentially expressed in scleroderma fibroblasts [abstract]. Arthritis Rheum 43(suppl):S259, 2000 Rogge L, Bianchi E, Biffi M, et al: Transcript imaging of the development of human T helper cells using oligonucleotide arrays. Nat Genet 25:96–101, 2000 Rus V, Luzina IG, Atamas SP, et al: DNA microarray analysis of cytokine and chemokine-related genes in peripheral blood mononuclear cells from lupus patients [abstract]. Arthritis Rheum 43(suppl):S95, 2000 Schena M, Shalon D, Davis RW, et al: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270:467–470, 1995 Scherf U, Ross DT, Waltham M, et al: A gene expression database for the molecular pharmacology of cancer. Nat Genet 24:236–244, 2000 Schroder AE, Greiner A, Seyfert C, et al: Differentiation of B cells in the nonlymphoid tissue of the synovial membrane of patients with rheumatoid arthritis. Proc Natl Acad Sci USA 93:221–225, 1996 Schummer M, Ng WV, Bumgarner RE, et al: Comparative hybridization of an array of 21,500 ovarian cDNAs for the discovery of genes overexpressed in ovarian carcinomas. Gene 238:375–385, 1999 Sherlock G: Analysis of large-scale gene expression data. Curr Opin Immunol 12:201– 205, 2000 Sherlock G, Hernandez-Boussard T, Kasarskis A, et al: The Stanford Microarray Database. Nucleic Acids Res 29:152–155, 2001 Shim C, Zhang W, Rhee CH, et al: Profiling of differentially expressed genes in human primary cervical cancer by complementary DNA expression array. Clin Cancer Res 4:3045–3050, 1998 Southern E, Mir K, Shchepinov M: Molecular interactions on microarrays. Nat Genet 21:5–9, 1999 Staudt LM, Brown PO: Genomic views of the immune system. Annu Rev Immunol 18:829–859, 2000 Stuhlmu¨ller B, Ungethu¨m U, Scholze S, et al: Identification of known and novel genes in activated monocytes from patients with rheumatoid arthritis. Arthritis Rheum 43:775–790, 2000 Tamayo P, Slonim D, Mesirov J, et al: Interpreting patterns of gene expression with self-organizing maps: Methods and application to hematopoietic differentiation. Proc Natl Acad Sci USA 96:2907–2912, 1999 Taniguchi K, Kohsaka H, Inoue N, et al: Induction of the p16INK4a senescence gene
176
74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85.
MOLITOR et al
as a new therapeutic strategy for the treatment of rheumatoid arthritis. Nat Med 5:760–767, 1999 Tavazoie S, Hughes JD, Campbell MJ, et al: Systematic determination of genetic network architecture. Nat Genet 22:281–285, 1999 Teague TK, Hildeman D, Kedl RM, et al: Activation changes the spectrum but not the diversity of genes expressed by T cells. Proc Natl Acad Sci USA 96:12691–12696, 1999 Trinchieri G: Interleukin-12: A proinflammatory cytokine with immunoregulatory functions that bridge innate resistance and antigen-specific adaptive immunity. Annu Rev Immunol 13:251–276, 1995 Verweij CL, Pouw van der Kraan CTM, Alizadeh AA, et al: Gene expression profiling of rheumatoid synovitis using cDNA microarrays [abstract]. Arthritis Rheum 43(suppl):S160, 2000 Wang K, Gan L, Jeffery E, et al: Monitoring gene expression profile changes in ovarian carcinomas using cDNA microarray. Gene 229:101–108, 1999 Wang T, Hopkins D, Schmidt C, et al: Identification of genes differentially overexpressed in lung squamous cell carcinoma using combination of cDNA subtraction and microarray analysis. Oncogene 19:1519–1528, 2000 Wellmann A, Thieblemont C, Pittaluga S, et al: Detection of differentially expressed genes in lymphomas using cDNA arrays: Identification of clusterin as a new diagnostic marker for anaplastic large-cell lymphomas. Blood 96:398–404, 2000 Woolf PJ, Wang Y: A fuzzy logic approach to analyzing gene expression data. Physiol Genomics 3:9–15, 2000 Young CL, Adamson TC, Vaughan JH, et al: Immunohistologic characterization of synovial membrane lymphocytes in rheumatoid arthritis. Arthritis Rheum 27:32–39, 1984 Zhou X, Tan FK, Xiong M, et al: Comparisons of fibroblasts gene expression profiles in systemic sclerosis, normal controls, scleromyxedema and eosinophilic fasciitis. Arthritis Rheum [abstract] 43(suppl):S259, 2000 Zhou X, Tan FK, Xiong M, et al: Expression of autoantigen genes is selectively and specifically altered in dermal fibroblasts of system sclerosis patients [abstract]. Arthritis Rheum 43(suppl):S259, 2000 Zhu H, Cong J-P, Mamtora G, et al: Cellular gene expression altered by human cytomegalovirus: Global monitoring with oligonucleotide arrays. Proc Natl Acad Sci USA 95:14470–14475, 1998 Address reprint requests to Gerald T. Nepom, MD, PhD Virginia Mason Research Center 1201 Ninth Avenue Seattle, WA 98101 e-mail:
[email protected]