NEW YORK (GenomeWeb News) – A new study is revealing the stickiness of transcription factors, and its authors said should remind researchers of both the wealth and limits of genome-wide approaches to understanding molecular interactions.
Researchers from the Lawrence Berkeley National Laboratory and the Santa Clara, California-based company Affymetrix found more than they bargained for while looking for genomic DNA associated with transcription factors involved in fruit fly development. The six proteins they tested bound extensively — several thousand places on the genome. But much of the binding fell outside of sites believed to influence gene expression.
The authors claim these results, published in PLoS Biology online today, highlight the importance of careful analysis of interaction data. “Our conclusions differ from those of other groups who have not distinguished between different levels of DNA binding in vivo using similar assays and have generally assumed that all detected binding is functional,” the authors wrote.
The group looked for transcription factor binding sites using straightforward ChIP-chip experiments. First, they did chromatin immunoprecipitation using antibodies to six well-characterized transcription factors acting in the early Drosophila melanogaster embryo. Then they applied these to Affymetrix whole-genome tiling arrays.
When it came to analyzing the data, though, they did something a bit different. Instead of classifying stretches of DNA as bound or unbound, they analyzed all of the interactions, classifying them based on their strength. Then, they mapped the location and properties of the genomic DNA associated with the six transcription factors at each level.
Not surprisingly, the strongest interactions were between the transcription factors and regions of DNA previously shown to functionally interact with the proteins. These “made a lot of sense in terms of what we already knew,” Lawrence Berkeley National Laboratory genomics researcher Mark Biggin, co-senior investigator on the paper, told GenomeWeb Daily News.
On the other hand, many of the weakly binding regions of the genome were less well conserved and didn’t map to regions where transcription factors would be expected. Some even fell within protein-coding regions of the genome.
“A shockingly high percentage of the weak binding regions are in protein-coding regions,” Biggin said. “It’s not at all consistent with anything people expect from a functional transcription factor binding site.”
This runs counter to the notion that more transcription factor binding sites correspond to more gene regulation. “I interpret this data with a little bit of caution,” Mike Eisen, a Lawrence Berkeley National Laboratory genomics researcher and the other co-senior investigator, told GenomeWeb Daily News. “In all likelihood, that binding isn’t doing anything.”
Although weak interactions aren’t necessarily non-functional, Biggin said, they do come under suspicion. But, he added, this information may add to the overall picture of what’s happening in the cell. Consequently, he believes molecular biologists should think about network interactions as a quantitative continuum rather than considering regions of the genome as “bound” or “unbound.”
“Our biggest message is that the data should be analyzed quantitatively,” Biggin said. “There is no magic threshold.”
Still, both Biggen and Eisen emphasized the importance of genome-wide approaches such as ChIP-chip — as long as the results are evaluated soberly. “The [genomic] method is profoundly useful. It’s just not — as with many things in life — enough,” Biggin said.