Just as CDs replaced cassette tapes, and MP3 music files are replacing CDs, DNA microarrays will eventually be obsolete. Genome-wide applications of microarrays are costly and limited to species with sequenced genomes. What will replace microarrays? New high-throughput sequencing platforms generate similar data and can be applied to any species. Johnson et al. replace DNA microarrays with high-throughput sequencing in a common technique that maps protein–DNA interactions in a recent article in Science.
Chromatin immunoprecipitation (CHiP) identifies protein–DNA interactions. Researchers crosslink endogenous proteins to bound DNA, lyse cells and isolate the protein–DNA complexes (chromatin). Then they digest the DNA into smaller units and precipitate the protein of interest along with its bound DNA with a specific antibody. After reversing the crosslink, they analyze the DNA sequences. For genome-wide CHiP application, researchers commonly analyze DNA sequences with microarrays in a technique called CHiP-chip.
Neuron-restrictive silencer factor (NRSF, also called repressor element-1 silencing transcription factor or REST) suppresses expression of neuronal genes in nonneuronal cells and neuronal progenitors. The classical NRSF-binding site contains 2 10-nucleotide half-sites whose centers are separated by 11 nucleotides. Many NRSF-binding sites have previously been identified by CHiP followed by quantitative PCR.
The authors immunoprecipitated NRSF and NRSF-bound DNA with an anti-NRSF antibody, and then sequenced several million 25-nucleotide DNA segments, with a technique they called CHiPSeq. They determined the location for each DNA segment within the genome and discarded regions detected less than 13 times. In all, they identified 1946 NRSF-binding sites near 1020 genes. Genes with NRSF-binding sites were enriched in gene ontology terms related to neurons, including synaptic transmission and nervous system development.
The authors compared NRSF-binding sites identified by CHiPSeq and CHiP-PCR. They graphed sensitivity (fraction of true positives) on the y-axis and specificity (fraction of false negatives) on the x-axis and examined the resulting receiver operator characteristic (ROC) curve. If all true positives and no false positives are detected, the ROC curve will have a y-axis value of 1.0 for all values of x and an area under the curve of 1.0. In contrast, random chance would produce a ROC curve with area under the curve of 0.5. CHiPSeq produced ROC curves with area equaling 0.96, suggesting that it is both sensitive and specific.
CHiPSeq identified new non-canonical NRSF-binding sites with 10 or 16-19 nucleotides between half-sites. The non-classical sites accounted for 197 NRSF-binding regions. The authors also identified new genes with NRSF-binding sites, including NEUROD1, which is involved in neuronal differentiation, and the hepatocyte nuclear factors HNF4a, HNF6 and Hes1.
Because one run of the sequencer can produce approximately 40 million reads of 25-nucleotide sequences, CHiPSeq might help map differences in transcription factor-binding sites from person-to-person.