Introduction

Complex transcriptional programs elicited by nuclear factor binding to regulatory elements are key processes in cellular systems, including macrophages (1). In silico computational approaches and in vitro assays have been widely used to determine the DNA binding characteristics of transcription factors. However, DNA binding sites are often highly flexible, and a purely bioinformatic prediction of motifs often lacks specificity, yielding a large number of false positives. Moreover, in vitro techniques such as electrophoretic mobility shift assays are poor predictors of actual binding in the nucleus. In vivo DNA footprinting and ChIP-PCR have overcome these problems for a small number of selected genomic loci but do not allow the identification of novel targets for transcription factors (2,3).

The combination of chromatin immunoprecipitation with DNA microarrays (ChIP-chip) is an emerging tool to generate genomewide maps of in vivo DNA-protein interactions (4,5). Successful ChIP-chip experiments principally rely on three important factors: (i) experimental optimization of immunoprecipitation for the given nuclear protein; (ii) DNA labeling, microarray hybridization, and processing; and (iii) data analysis and sequence identification at the genome level (6). Established protocols for chromatin immunoprecipitation with pioneering work from the Farnham laboratory (7,8) and standardized microarrays (9,10) helped to increase the reproducibility and minimize the experimental bias. Nonetheless, in contrast to gene expression micrarrays, ChIP-chip raw data analysis still requires several laborious and complicated bioinformatic steps and several programs to generate useful biological information. Thus, analysis of promoter tiling arrays initially requires algorithms for normalization of single probe intensities and calculation of signals and P values. After defining thresholds and window sizes, logarithmic ratios of transcription factor–enriched versus non-enriched genomic regions are used to identify positive chromosomal intervals (peaks). Independent programs and genomic databases are then used for annotation and identification of neighboring genes. Given that well-characterized nucleotide matrices are available for the transcription factor, in silico analyses can be performed to narrow and allocate putative binding sites. Moreover, the sequences of identified target regions can be used to generate binding matrices.

Although several programs have been described that cover single parts of these analysis steps, software packages combining all key features of ChIP-chip results are rare. We explored whether we could successfully implement a single workflow that integrates all steps of microarray analysis, identification of genomic regions, correlation with transcription factor binding sites, and gene network analysis. As an example, we report here the first genome-wide identification of target promoters for the hematopoietic transcription factor PU.1 in macrophages using the Affymetrix GeneChip Mouse Promoter 1.0R arrays (Affymetrix, Santa Clara, CA, USA) and the ChipInspector software tool (Genomatix GmbH, Munich, Germany).

Materials and Methods

ChIP-chip Assay

Chromatin immunoprecipitation from RAW264.7 cells was performed as described previously (11). Briefly, 10⁶ RAW264.7 cells were cross-linked for 10 min with formaldehyde, nuclei were lysed, and the chromatin was prepared by sonication. The lysates were precleared with 5 µg salmon sperm DNA/sepharose CL-4B beads and precipitated with 2.5 µg polyclonal PU.1 antibody or IgG rabbit isotype control (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Complexes were recovered with protein A-/protein G-sepharose beads (Upstate, Lake Placid, NY, USA) and cross-linking was reversed. The DNA was purified with QIAquick PCR purification columns (Qiagen, Hilden, Germany) and enrichment was analyzed by quantitative real-time PCR with primers amplifying the known PU.1 target promoter Dap12 (forward, 5′-TCAAGGCCCAGAGAAGCTAA-3′; reverse, 5′-CATGAGCTGAGGACACAG-3′).

Three biological replicates for PU.1 and IgG were generated and successful enrichment was confirmed in each experiment. For ChIP-chip, PU.1 and IgG-immunoprecipitated DNA were amplified, fragmented, and labeled according to the Affymetrix ChIP Assay Protocol (www.affymetrix.com/support/downloads/manuals/chromatin_immun_ChIP.pdf). Seven and a half micrograms of ∼500 bp DNA fragments were hybridized to six Mouse Promoter 1.0R GeneChips (Affymetrix). Each single array interrogated more than 25,000 promoters with a coverage of 6 kb upstream and 2.5 kb downstream of transcription start sites. The 25-mer oligonucleotide probes were tiled at a resolution of 35 bp with gaps of 10 bp, allowing the identification of enriched elements with an optimal resolution and specificity. After scanning, the cell intensities of each feature were computed with the GeneChip Operating Software (GCOS; Affymetrix) to obtain CEL files.