PERSPECTIVE
Genomic Microarrays
Arraying Order in Biological Chaos?

Rafeul Alam and Magdalena Gorska

University of Texas Medical Branch, Department of Internal Medicine, Division of Allergy & Immunology, Galveston, Texas

The recent sequencing of the human genome has opened a new era in biomedical research. The main question now is how to use all the data generated by Human Genome Project and convert DNA sequence into the cellular function. The main effort is now being placed on characterization of gene products, understanding their role in physiology, and delineating their association with human diseases. DNA microarrays are among the techniques which can help to answer these questions (reviewed in (1)).

	Technology

DNA microarrays are used to estimate the levels of mRNA in the cell. The technique is based on the hybridization of the fluorescently labeled complementary DNA/cDNA probes made from total cellular messenger RNA (mRNA) to an array of DNA representing human genes (1, 2). After the hybridization the slide is scanned by laser. Laser causes excitation of fluorescently labeled cDNAs probes. Only the spots representing mRNAs in the sample give emission signals. The emission is measured using a scanning confocal laser microscope and data are analyzed by appropriate software. The absence of the fluorescence of the specific spot means that complementary mRNA is not present in the sample. If the fluorescence is present, the intensity of the signal is a measure of the level of particular mRNA in the examined cell population. Usually an investigator wants to compare mRNA abundance in two different samples (or a sample and a control). In this case, cDNAs from the sample and a control are labeled with two different fluorescent dyes (e.g., red label for the cDNAs from the sample and a green label for the control). Next these two cDNA populations are allowed to hybridize to the same microarray slide. If particular mRNA from the sample is in abundance, the spot with a complementary probe will be red; if the concentration of the particular mRNA is higher in the control, the spot will be green. If both samples contain the same amount of a given mRNA, the spot will be yellow. Thus one can conclude about the relative expression levels of the genes for colors and fluorescent intensities of the microarray spots.

	Expectation

The general expectation has been that microarray experiments will give us a holistic view of biology as opposed to the fragmented reductionist view provided by the currently dominated experimental approach. However, this expectation may be overly optimistic since genomic microarrays give us the view of inducible genes only. Many biological processes occur without the participation of newly induced genes. These processes simply engage the existing protein pool with or without their posttranslational modification and may actually precede the transcription of new genes. Examples include cell migration and secretion of preformed mediators. To gain a true holistic view of biological processes we must, therefore, combine the genomic microarray with the proteomic microarray. For now, genomic microarrays are used for the following purposes:

Determination of transcriptional programs of cells for a given cellular function (e.g., cell cycle, cell differentiation, etc.) or when they are exposed to certain conditions leading to activation, inhibition, or apoptosis. Most current investigations focus on this type of research. These experiments allow identification of novel proteins involved in particular cellular processes. Further, they enable scientists to create a database of a complete transcriptional program for a given cellular function. For example, a complete transcriptional program involved in the yeast cell cycle has been delineated (4). In this issue, Temple and colleagues present data on the transcriptional program of eosinophils that are stimulated with interleukin (IL)-5 (5). The authors show that a number of genes are regulated by more than three-fold. One of the goals of this study was to identify genes involved in regulating eosinophil survival and apoptosis. Since IL-5 not only delays eosinophil apoptosis but also activates the cell for numerous functions, it is not clear whether the upregulated genes are directly involved in eosinophil survival. For this reason the authors expanded their study to include the TF-1 cell line, which depends on IL-5 as a growth and survival factor. They compared the upregulated genes from IL-5-stimulated eosinophils with the downregulated genes from TF-1 cells that were deprived of IL-5. Using this comparison the authors show that four genes are regulated by IL-5 in a coordinated manner in both cell types (i.e., eosinophils and TF-1 cells). These genes include Pim-1, SLP-76, CD24, and DSP-5. Although the authors do not present any data to support the role of these gene products in eosinophil survival, they provide a compelling argument to support their conclusion (see discussion below).

Compare and contrast transcriptional programs to aid diagnosis and prognosis of diseases, and predict therapeutic response. This is truly an exciting field with an immediate impact on disease diagnosis and appropriate treatment. Classification of many diseases including cancer and rheumatoid diseases are frequently based upon ill-defined and overlapping criteria. Comparing and contrasting transcriptional programs may help develop better criteria for disease classification and predict therapeutic response. An example is the difficulty in the diagnosis of small round blue cell tumors (SRBCT) of childhood (6). SRBCT include neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, and the Ewing's family of tumors. These tumors are difficult to diagnose by routine histology and require several additional techniques including histochemistry, cytogenetics, and interphase fluorescence in situ hybridization. By examining the transcriptional program of these tumors and analyzing the data by artificial neural networks (computer-based algorithms modeled after the structure and behavior of neurons), a new classification of SRBCT has been developed, which promises to be most comprehensive and clinically useful. Another example is the surprising observation that diffuse B cell lymphoma is not a single disease (7). Microarray analysis suggests that this clinical condition actually comprises two separate diseases with two distinct transcription profiles. Patients with this disease vary in their response to therapy and the transcription profile-based classification may explain this variable response to treatment. A similar approach could prove useful in diagnosing a variety of conditions including many overlapping rheumatologic conditions.

Identification of genome-wide binding sites for transcription factors. This is a novel application of the microarray technology. The transcription of genes is regulated by binding of transcription factors to the promoter region. Until now, transcription factor binding sites are identified on a single gene basis. This novel application of microarrays will allow identification of genome-wide binding sites. The basic principle is to crosslink genomic DNA to a given transcription factor in situ, isolate protein-bound DNA pieces by immunoprecipitation, amplify and label the DNA, and finally hybridize the probes to microarrays representing putative promotersfor example, all of the intergenic, noncoding sequences (8). The above method was successfully used to identify nearly 200 new binding sites for the yeast transcription factors SBF and MBF.

Prediction of gene function. In addition to the previously identified and characterized genes, microarray experiments frequently show transcription of novel genes, whose functions are unknown. One of the challenges of microarray experimentation is to predict the function of novel genes based upon theoretical modeling. One way to accomplish this goal is to cluster genes according to their expression profiles that are generated using a range of conditions. A cluster of genes will be assigned a given functional class based upon the function of the majority of the genes in the cluster. New genes falling in a functional cluster are predicted to have a similar function. This guilt-by-association approach has been shown useful in predicting novel gene function in the yeast (9).

Identification of new therapeutic targets. This objective is quite obvious given that the microarray technology allows identification of disease-associated genes (not necessarily disease-causing genes) in a comprehensive manner. Genomic microarray is being integrated into many steps of drug development including target identification, target validation, and drug toxicity (reviewed in ref. (3)).

Development of public databases. The development of public databases will help us understand the functioning of complex biological systems, which has been evading our quest until now. The combination of biology, computer science, and mathematics hopefully will result in the ability not only to analyze but also to build models. Ultimately it will help us develop "virtual cells" and "virtual organisms."

	Challenges

Microarray experiments usually produce a huge amount of data. The biggest challenge at this time is data mining and interpretation. Stimulation of cells induces the transcription of molecules belonging to diverse families, which may include transcription factors, cytokines, receptors and ion channels, signaling molecules, cytoskeletal molecules, metabolic and trafficking molecules, and their regulators. The challenge is to understand the role of all inducible genes in defining the phenotype of a cell. Do all the induced genes contribute to the phenotype or are some genes unintentionally transcribed because they share common transcriptional factors (a by-stander effect)? This important question needs to be addressed before we fully understand the relationship between a transcriptional program and a specific cellular phenotype.

	Surprises

Since microarray provides a complete transcriptional program for a given stimulus, expectation for surprises has been high. A few surprises are worth mentioning. In a study of Hodgkin lymphoma unexpectedly the production of IL-5 and IL-13 is found to be significantly elevated compared to large cell lymphoma and other non-Hodgkin lymphoma (10). Anti-IL-13 antibody blocks proliferation of Hodgkin cells. This is quite exciting since this microarray study identifies an excellent therapeutic target. Humanized anti-IL-13 antibody or a receptor antagonist may become therapeutically useful in Hodgkin lymphoma. Another surprise comes from a microarray study of atherosclerotic lesions. This study demonstrates significant upregulation of eotaxin and the CC chemokine receptor-3 (CCR3) in atherosclerotic lesions (11). Although the biological relevance of this finding is unclear, nonetheless, it provides a rationale to study the effect of CCR-3 antagonists on atherosclerosis development. Alternatively, CCR-3 knockout mice can be studied to explore this subject. Another microarray study shows an unexpected production of large quantities of plasminogen activator inhibitor-1 (PAI-1) by mast cells (12). The plasminogen activator system plays a crucial role in controlling extracellular matrix proteolysis leading to tissue remodeling and fibrosis. PAI-1-deficient mice are resistant to pulmonary fibrosis. Thus, the production of PAI-1 points to a pathogenic role of mast cells in airway remodeling in addition to its known role in asthma.

The microarray experiment by Temple and coworkers reveals that IL-5 regulates an interesting group of genes (5). The regulation of some of the genes such as CD69, CCR1, and IL-8, has previously been reported (reviewed in (13)) and was, therefore, anticipated. The new finding includes the regulation of Pim-1, SLP-76, CD24, and DSP-5. CD24 is an adhesion molecule and its cross-linking induces B cell apoptosis (14). Its upregulation by IL-5 in eosinophils is intriguing and will require further studies. The role of DSP-5, a phosphatase, in eosinophil survival is unclear. The most interesting findings relate to upregulation of Pim-1 and SLP-76. Pim-1 is a serine-threonine kinase, which has been shown to be important for hematopoietic growth factor-induced myeloid cell growth and survival (15, 16). Pim-1 regulates gene transcription by phosphorylating heterochromatin protein-1 (HP-1) (17). It also phosphorylates PTP-U2S, a tyrosine phosphatase (18). It interacts with c-myc family of transcription factors (19) and upregulates Bcl-2 family members (20). Strong induction of SLP-76 by IL-5 in eosinophils is another interesting finding. SLP-76 is an adapter protein, which upon phosphorylation by tyrosine kinases transduces downstream signaling via Grb2, PLC-, Vav, and SLAP-130 (Fyb) (21). It plays an important role in signaling through the T cell antigen receptor (22) and mast cell FcRI (23). The importance of SLP-76 for eosinophil growth and survival has not been previously investigated. If future studies confirm that Pim-1 and SLP-76 are critical for eosinophil survival and activation, the two molecules could become excellent targets for interference with IL-5 signaling.

Previous studies have shown that two receptor-associated tyrosine kinasesJak2 and Lynplay a crucial role in IL-5 prolongation of eosinophil survival (24, 25). It is interesting to note that Lyn kinase is linked to SLP-76 (26), whereas multiple signaling pathways including the Jak-Stat, phosphatidyl inositol-3 kinase, and the MAP kinase pathways regulate Pim-1 (27). SLP-76 is likely to be important for transducing signals via Raf-1 and the MAP kinase pathway. Raf-1 has been identified as critical survival regulator for eosinophils (24). PI-3 kinase also regulates myeloid cell survival through c-Akt (28). Both c-Akt and Raf-1 interact with Bcl-2 family members, which are the direct regulators of apoptotic pathways (29, 30). Since Pim-1 also upregulates Bcl-2 family members, the data suggest that many survival signals ultimately converge on the Bcl-2 family (Figure 1). The important question is why these diverse signaling pathways, which engage many other signaling molecules, preferentially increase the transcription of Pim-1 and SLP-76. One possibility is that these two signaling molecules have a rate-limiting role for eosinophil survival and their synthesis is downregulated by proapoptotic factors. Clearly, we will need more studies in order to fully understand the findings of Temple and colleagues. Nonetheless, the study has already identified transcriptionally regulated key signaling molecules that determine the survival fate of eosinophils. This is just the beginning of excitement.

View larger version (29K): [in this window] [in a new window]   Figure 1.   IL-5-activated signaling pathways leading to eosinophil survival and activation.

	Footnotes

Address correspondence to: Rafeul Alam, M.D., Ph.D., The University of Texas Medical Branch, Dept. of Internal Medicine, Division of Allergy & Immunology, Galveston, TX 77555-1083. E-mail: ralam{at}utmb.edu

(Received in original form August 27, 2001).

	References

1. Celis, J. E., M. Kruhoffer, I. Gromova, C. Frederiksen, M. Ostergaard, T. Thykjaer, P. Gromov, J. Yu, H. Palsdottir, N. Magnusson, and T. F. Orntoft. 2000. Gene expression profiling: monitoring transcription and translation products using DNA microarrays and proteomics. FEBS Lett. 480: 2-16 [Medline].

2. Brown, P. O., and D. Botstein. 1999. Exploring the new world of the genome with DNA microarrays. Nat. Genet. 21(1Suppl.):33-37.

3. Young, R. A.. 2000. Biomedical discovery with DNA arrays. Cell 102: 9-15 [Medline].

4. Cho, R. J., M. J. Campbell, E. A. Winzeler, L. Steinmetz, A. Conway, L. Wodicka, T. G. Wolfsberg, A. E. Gabrielian, D. Landsman, D. J. Lockhart, and R. W. Davis. 1998. A genome-wide transcriptional analysis of the mitotic cell cycle. Mol. Cell 2: 65-73 . [Medline]

5. Temple, R., E. Allen, J. Fordman, S. Phipps, H.-C. Schneider, K. Kindauer, I. Hayes, J. Lockey, K. Pollock, and R. Jupp. 2001. Microarray analysis of eosinophils reveals a number of candidate survival and apoptosis genes. Am. J. Respir. Cell Mol. Biol. 25: 425-433 [Abstract/Free Full Text].

6. Khan, J., J. S. Wei, M. Ringner, L. H. Saal, M. Ladanyi, F. Westermann, F. Berthold, M. Schwab, C. R. Antonescu, C. Peterson, and P. S. Meltzer. 2001. Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks. Nat. Med. 7: 673-679 [Medline].

7. Alizadeh, A. A., M. B. Eisen, R. E. Davis, C. Ma, I. S. Lossos, A. Rosenwald, J. C. Boldrick, H. Sabet, T. Tran, X. Yu, J. I. Powell, L. Yang, G. E. Marti, T. Moore, J. Hudson Jr., L. Lu, D. B. Lewis, R. Tibshirani, G. Sherlock, W. C. Chan, T. C. Greiner, D. D. Weisenburger, J. O. Armitage, R. Warnke, L. M. Staudt, and et al. 2000. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 403: 491-492 [Medline].

8. Iyer, V. R., C. E. Horak, C. S. Scafe, D. Botstein, M. Snyder, and P. O. Brown. 2001. Genomic binding sites of the yeast cell-cycle transcription factors SBF and MBF. Nature 409: 533-538 [Medline].

9. Hughes, T. R., M. J. Marton, A. R. Jones, C. J. Roberts, R. Stoughton, C. D. Armour, H. A. Bennett, E. Coffey, H. Dai, Y. D. He, M. J. Kidd, A. M. King, M. R. Meyer, D. Slade, P. Y. Lum, S. B. Stepaniants, D. D. Shoemaker, D. Gachotte, K. Chakraburtty, J. Simon, M. Bard, and S. H. Friend. 2000. Functional discovery via a compendium of expression profiles. Cell 102: 109-126 [Medline].

10. Kapp, U., W. C. Yeh, B. Patterson, A. J. Elia, D. Kagi, A. Ho, A. Hessel, M. Tipsword, A. Williams, C. Mirtsos, A. Itie, M. Moyle, and T. W. Mak. 1999. Interleukin 13 is secreted by and stimulates the growth of the Hodgkin and Reed-Sterberg Cells. J. Exp. Med. 189: 1939-1946 [Abstract/Free Full Text].

11. Haley, K. J., C. M. Lilly, J. H. Yang, Y. Feng, S. P. Kennedy, T. G. Turi, J. F. Thompson, G. H. Sukhova, P. Libby, and R. T. Lee. 2000. Overexpression of eotaxin and the CCR3 receptor in human atherosclerosis: using genomic technology to identify a potential novel pathway of vascular inflammation. Circulation 102: 2185-2189 [Abstract/Free Full Text].

12. Cho, S. H., S. W. Tam, S. Demissie-Sanders, S. A. Filler, and C. K. Oh. 2000. Production of plasminogen activator inhibitor-1 by human mast cells and its possible role in asthma. J. Immunol. 165: 3154-3161 [Abstract/Free Full Text].

13. Hartnell, A., D. S. Robinson, A. B. Kay, and A. J. Wardlaw. 1993. CD69 is expressed by human eosinophils activated in vivo in asthma and in vitro by cytokines. Immunology 80: 281-286 [Medline].

14. Suzuki, T., N. Kiyokawa, T. Taguchi, T. Sekino, Y. U. Katagiri, and J. Fujimoto. 2001. CD24 induces apoptosis in human B cells via the glycolipid-enriched membrane domains/rafts-mediated signaling system. J. Immunol. 166: 5567-5577 [Abstract/Free Full Text].

15. Pircher, T. J., S. Zhao, J. N. Geiger, B. Joneja, and D. M. Wojchowski. 2000. Pim-1 kinase protects hematopoietic FDC cells from genotoxin-induced death. Oncogene 19: 3684-3692 [Medline].

16. Shirogane, T., T. Fukada, J. M. Muller, D. T. Shima, M. Hibi, and T. Hirano. 1999. Synergistic roles for Pim-1 and c-Myc in STAT3-mediated cell cycle progression and antiapoptosis. Immunity 11: 709-719 [Medline].

17. Koike, N., H. Maita, T. Taira, H. Ariga, and S. M. Iguchi-Ariga. 2000. Identification of heterochromatin protein 1 (HP1) as a phosphorylation target by Pim-1 kinase and the effect of phosphorylation on the transcriptional repression function of HP1(1). FEBS Lett. 467: 17-21 [Medline].

18. Wang, Z., N. Bhattacharya, M. K. Meyer, H. Seimiya, T. Tsuruo, J. A. Tonani, and N. S. Magnuson. 2001. Pim-1 negatively regulates the activity of PTP-U2S phosphatase and influences terminal differentiation and apoptosis of monoblastoid leukemia cells. Arch. Biochem. Biophys 390: 9-18 [Medline].

19. Mochizuki, T., C. Kitanaka, K. Noguchi, T. Muramatsu, A. Asai, and Y. Kuchino. 1999. Physical and functional interactions between Pim-1 kinase and Cdc25A phosphatase. Implications for the Pim-1-mediated activation of the c-Myc signaling pathway. J. Biol. Chem. 274: 18659-18666 [Abstract/Free Full Text].

20. Lilly, M., J. Sandholm, J. J. Cooper, P. J. Koskinen, and A. Kraft. 1999. The PIM-1 serine kinase prolongs survival and inhibits apoptosis-related mitochondrial dysfunction in part through a bcl-2-dependent pathway. Oncogene 18: 4022-4031 [Medline].

21. Pivniouk, V. I., and R. S. Geha. 2000. The role of SLP-76 and LAT in lymphocyte development. Curr. Opin. Immunol. 12: 173-178 [Medline].

22. Pivniouk, V., E. Tsitsikov, P. Swinton, G. Rathbun, F. W. Alt, and R. S. Geha. 1998. Impaired viability and profound block in thymocyte development in mice lacking the adaptor protein SLP-76. Cell 94: 229-238 [Medline].

23. Pivniouk, V. I., T. R. Martin, J. M. Lu-Kuo, H. R. Katz, H. C. Oettgen, and R. S. Geha. 1999. SLP-76 deficiency impairs signaling via the high-affinity IgE receptor in mast cells. J. Clin. Invest. 103: 1737-1743 [Medline].

24. Pazdrak, P., B. Olszewska-Pazdrak, S. Stafford, and R. Alam. 1998. Lyn, Jak2 and Raf-1 kinases are critical for the anti-apoptotic effect of interleukin-5 whereas only Raf-1 kinase is essential for eosinophil activation and degranulation. J. Exp. Med. 188: 421-429 [Abstract/Free Full Text].

25. Yousefi, S., D. C. Hoessli, K. Blaser, G. B. Mills, and H. U. Simon. 1996. Requirement of Lyn and Syk tyrosine kinases for the prevention of apoptosis by cytokines in human eosinophils. J. Exp. Med. 183: 1407-1714 [Abstract/Free Full Text].

26. Gross, B. S., S. K. Melford, and S. P. Watson. 1999. Evidence that phospholipase C-gamma2 interacts with SLP-76, Syk, Lyn, LAT and the Fc receptor gamma-chain after stimulation of the collagen receptor glycoprotein VI in human platelets. Eur. J. Biochem. 263: 612-623 [Medline].

27. Krumenacker, J. S., V. S. Narang, D. J. Buckley, and A. R. Buckley. 2001. Prolactin signaling to pim-1 expression: a role for phosphatidylinositol 3-kinase. J. Neuroimmunol. 113: 249-259 [Medline].

28. Eves, E. M., W. Xiong, A. Bellacosa, S. G. Kennedy, P. N. Tsichlis, M. R. Rosner, and N. Hay. 1998. Akt, a target of phosphatidylinositol 3-kinase, inhibits apoptosis in a differentiating neuronal cell line. Mol. Cell. Biol. 18: 2143-2152 [Abstract/Free Full Text].

29. Huang, H. M., C. J. Huang, and J. J. Yen. 2000. Mcl-1 is a common target of stem cell factor and interleukin-5 for apoptosis prevention activity via MEK/MAPK and PI-3K/Akt pathways. Blood 96: 1764-1771 [Abstract/Free Full Text].

30. Wang, H. G., U. R. Rapp, and J. C. Reed. 1996. Bcl-2 targets the protein kinase Raf-1 to mitochondria. Cell 87: 629-638 [Medline].