Identification of PML-RARa Target Genes Using Microarray and ChIP-on-Chip Analysis.

PML-RARa is very toxic to early myeloid cell lines that express neutrophil elastase. As a consequence, myeloid cell lines that stably express PML-RARa must undergo significant adaptations to avoid toxicity (eg. loss of expression of neutrophil elastase (Lane and Ley, MCB, 2005). To identify the direct target genes of PML-RARa, we decided to bypass the issue of adaptation by profiling naive U937 cells that transiently expressed PML-RARa. To define “immediate-early” expression changes caused by PML-RARa, we harvested U937 cells electroporated with expression vectors containing eGFP-PML-RARa (known to possess all the measurable functional properties of PML-RARa) vs. eGFP alone, at the earliest time point when adequate numbers of eGFP positive cells could be collected via high speed cell sorting (6 hours). RNA obtained from these cells was subjected to linear amplification and hybridized with Affymetrix U133 Plus 2.0 arrays. Probesets that demonstrated the largest differences in expression between cells transfected with eGFP vs. eGFP-PML-RARa were PML and RARa themselves (15-fold increase for PML, 9-fold for RARa), providing an important validation for the experimental system. Differentially expressed genes were selected by three filters for further study: (i) ANOVA statistic (P < 0.05); (ii) “present” call and >150 average expression units in at least one sample; and (iii) Reproducible (n=2) >2-fold differences between eGFP-PML-RARa vs. eGFP vector transfected cells. Only 280 probesets (out of 54,613 total) met all of these criteria. Among them, 37 genes were involved in signal transduction, 15 in cell proliferation and differentiation, 7 in apoptosis, 14 in cell cycle, 15 in transport, 27 in protein metabolism and modification, 13 in oncogenesis, and 13 in immunity and defense (Panther classification system). 15/15 genes selected from a variety of categories were further validated with Q-PCR in duplicate using replicate samples, and all differences were statistically significant (p<0.05). Q-PCR validated genes that were up-regulated at 6 hours by eGFP-PML-RARa included CD36 (196% increase over eGFP vector control), LATS2 (204%), Sox4 (395%), COPEB (325%), HoxA5 (294%), IFITM3 (139%), AIF1 (197%), and PLD1 (315%). Q-PCR validated genes that were down-regulated by eGFP-PML-RARa at 6 hours included BTG2 (60% the level of eGFP vector control), KLF7(52%), CDKN1A (p21)(39%), Pim1 (58%), SLC23A2 (36%), HoxA13 (51%), and PDCD1LG1 (42%). Among these genes, Sox4, Pim1, BTG2, and PDCD1LG1 have previously been implicated in cancer pathogenesis by retroviral insertional mutagenesis (http://RTCGD.ncifcrf.gov). To identify promoters that were directly in contact with the eGFP-PML-RARa fusion protein, we performed “ChIP-on-chip” analysis using chromatin immunoprecipitation coupled with a NimbleGen promoter tiling array containing 15 50-mer probes (~100 bp spacing) between positions −1500 and +500 of the transcription initiation sites of a curated subset of 24,275 known human genes (HG17, Build 35). By subtracting non-specific hybridization signals from the eGFP control chip, we identified and validated several promoter regions as direct binding targets of eGFP-PML-RARa, including HoxA13 and CDKN1A (p21). This experimental system provides a new approach for the identification of the direct transcriptional targets of PML-RARa, which may lead to new insights into APL pathogenesis and treatment.