Transcriptional regulatory networks: Pitfalls and Surprises

The use of high-throughput methods to reconstruct the transcriptional regulatory program of a species is an exciting development of the post-genomic era. At least three distinct methods based on unrelated principles are in vogue and include reconstructions of the transcriptional program by 1) over-expression of transcription factors; 2) deletion of transcription factors; and 3) large-scale Chromatin immunoprecipitation-chip (ChIP-chip) experiments. From these data, a transcriptional regulatory network (TRN) can be constructed where a node represents a transcription factor (TF) or target gene (TG) and an edge is made between a TF and a TG if the TF has a regulatory effect on a TG, or if the TF binds the promoter of the TG.

Recently, we along with Dr. Madan Babu of MRC-UK, compared the transcriptional regulatory networks of the budding yeast, Saccharomyces cerevisiae, reconstructed from the above datasets and reached some surprising conclusions.

• Despite sharing a significant number of TFs, the overlap in the regulatory interactions between the networks is strikingly small (<5%).
• The network structure differs between the different TRN reconstructions. For example, the number of TGs regulated by a given TF shows a power-law distribution in the ChIP-chip and TF deletion networks. In contrast the same distribution in the TF over-expression network shows a central tendency.
• Although the ChIP-chip and TF deletion networks show a similar global structure, only about a quarter of the total number of hubs (top 20% of TFs with the greatest number of TGs) are shared between the two networks.
  
• We also show that unintended consequences of experimental design, may significantly influence the TRNs. For example, in the ChIP-chip network, telomeric looping effect, or the interaction of chromosome ends with diverse TFs in the inner nuclear envelope may have contributed to the unusually large number of TFs bound to the promoters of several subtelomeric TGs. An analysis of the gene over-expression TRN revealed that several TGs that are regulated by a large number of TFs were components of stress response pathway/s pertaining to protein unfolding and oxidative damage. One such set of proteins that are expressed are the homologs of human DJ1 implicated in protein aggregation defect in Parkinson's disease. Thus over-expression of TFs may cause an increase in mis-folded polypeptides triggering the expression of proteins involved in a specific stress response pathway. Similarly, in the gene deletion TRNs, most of the top hubs have regulatory interactions with ribosomal components. Thus altering the expression of these proteins may alter the stoichiometry of the ribosomal components which in turn may indirectly affect the expression of several genes.
  

Our study shows that each of these TRN reconstructions captures a different aspect of the transcriptional regulatory program, Secondary effects caused by experimental design may significantly influence the network and future high-throughput experiments must subtract such additional effects to obtain more accurate TRNs. Finally, high-throughput methods may throw unexpected and novel insights into specific biological phenomena.
For more details read the full article here.