Transcription factor binding dominates talks at minisymposium
By Jeffrey Stumpf
A cast of scientists at the forefront of 21st century techniques was on hand at NIEHS May 29-30 for a pair of half-day symposia. NIEHS researcher Leping Li, Ph.D., organized the nine talks on motif discovery and sequence data analysis that attracted renowned scientists from prestigious universities, as well as an NIEHS audience with wide-ranging interests. Li’s colleagues Pierre Bushel, Ph.D., of NIEHS, and Jason Lieb, Ph.D., (http://bio.unc.edu/people/faculty/lieb/) of the University of North Carolina at Chapel Hill, chaired the sessions.
Duke University researcher Greg Crawford, Ph.D., (http://www.genome.duke.edu/directory/faculty/crawford/) summarized the speakers’ interests in genome sciences by reminding the audience that 98 percent of the genome does not code for amino acid sequences. The meeting emphasized many ways that the noncoding regions are involved in gene regulation.
Recurring motifs describe nucleotide binding
Transcription factors bind to DNA at consensus sequences, or motifs, to regulate gene expression. Keynote speaker Gary Stormo, Ph.D., (http://dbbs.wustl.edu/faculty/Pages/faculty_bio.aspx?SID=3439) from Washington University in St. Louis, provided a multidisciplinary approach to understanding where transcription factors bind. Stormo used high-throughput experiments to measure the short DNA sequences to which transcription factors bind and promote changes in expression. These methods would ultimately be used to determine binding site motifs based on the amino acid sequence of the protein.
While DNA was the favored nucleotide at the symposium, Brown University computational biologist, Charles Lawrence, Ph.D., (http://www.dam.brown.edu/people/facultypage.lawrence.html) proudly provided a respite. “I promise, there will be no DNA in this talk,” joked Lawrence, who instead focused on gene regulation by microRNAs (miRNAs). An average of only 22 nucleotides long, miRNAs are prevalent in the human genome and repress gene expression by binding to transcripts. Because finding the target for miRNAs is difficult, Lawrence discussed how his mathematical models not only improve prediction of miRNA secondary structures, but also dissect a possible transcript-binding motif.
When the nucleosomes are away, the transcription factors play
Although DNA binding motifs are useful in discovering transcriptional networks, there are determinants for DNA binding other than sequence. In the course of a session that showcased how researchers took advantage of locating regions where nucleosomes bind DNA, several speakers emphasized that the binding of nucleosomes to DNA deters transcription factor activation.
“We can use nucleosome eviction as a marker to predict interesting regulatory activity,” Lieb explained.
Cliff Meyer, Ph.D., (http://liulab.dfci.harvard.edu/~cliff/) a researcher at the Harvard School of Public Health, likened the study of transcription factor binding to Plato’s famous example of inferring truth from shadows in his Allegory of the Cave. To ascend from the proverbial cave, Meyer identified regions that are hypersensitive to DNase treatment, indicating sections of the genome where nucleosomes bind less frequently.
“The upregulation of gene expression correlated with the lack of nucleosome occupancy globally and vice versa,” Meyer explained.
Lieb used a similar approach to determine nucleosome occupancy but applied it to Drosophila development. By searching regions with decreased nucleosome binding, Lieb identified a regulatory element in a 27-kilobase region that programs cells to become halteres, structures that improve balance during Drosophila flight. Furthermore, Lieb described chromatin profiles of various cell fates at different times, and was surprised by the conclusion.
“Regarding the chromatin profile, different tissues at the same stage in development are more similar than their own daughter cells,” Lieb summarized.
Computer modeling aims for usefulness
In light of the biological data presented by Lieb and Meyer, Duke University professor Alexander Hartemink, Ph.D., (http://www.cs.duke.edu/~amink/) provided a computer-generated model called COMPETE to visualize the competition between nucleosomes and transcription factors in binding DNA. His model demonstrated that increasing one transcription factor might displace nucleosomes to other regions, causing reduced binding of nearby transcription factors.
Considering his model was developed using only computational methods, Hartemink interpreted his results cautiously, reminding the audience of a quote from the statistician George Box, Ph.D., “All models are wrong, but some models are useful.”
(Jeffrey Stumpf, Ph.D., is a research fellow in the NIEHS Laboratory of Molecular Genetics Mitochondrial DNA Replication Group and a regular contributor to the Environmental Factor.)
Symposium highlights cross talk between biologists and computer scientists
The search for DNA binding motifs involves high-throughput techniques that only have been possible within the last decade. NIEHS Scientific Director Darryl Zeldin, M.D., introduced the symposium with the message that mining through the wealth of data is our next greatest challenge.
“As our ability to generate large amounts of genetic data has advanced at an astronomical rate, our ability to analyze the data and understand fundamental biological questions has not [advanced at the same pace],” Zeldin remarked.
Li also acknowledged the importance of mixing the best of two separate, yet equally, important worlds.
“Advances in technologies such as next-gen sequencing make it possible to study genome-wide gene expression or protein binding in a single experiment,” Li noted. “However, making sense of the huge amount of data remains a challenge, making bioinformatics an increasingly integral part of modern biology.”
The marriage of two distinct disciplines not only has shaped today’s research, but also has provided the science community with many computational tools. Many of these programs are publicly available, including ENCODE (the Encyclopedia of DNA Elements, produced by Crawford), BACH (Bayesian 3D constructor for Hi-C data, described by speaker Jun Liu, Ph.D., (http://www.hsph.harvard.edu/faculty/jun-liu/) of the Harvard School of Public Health), ChromHMM (Chromatin-state discovery and characterization, written by speaker Jason Ernst, Ph.D., (http://faculty.biolchem.ucla.edu/institution/personnel?personnel_id=2028381) of UCLA), and COMPETE.
Although the symposium featured speakers with diverse experience and expertise, Zeldin noted their common attributes.
“These are internationally recognized thought leaders who have achieved both successes — not only dealing with sophisticated biological questions, but also rendering the findings accessible to answer important environmental health problems,” Zeldin concluded.