Search for a command to run...
Systems biology seeks to understand the interconnections and interplay between the myriad of biological processes that occur within and between cells. Understanding this complex network of events requires a characterization of the molecular interactions that control even the simplest regulatory schemes. Acquisition of detailed DNA sequence information for a growing number of organisms should provide a means of probing interconnections between multiple regulatory events, including the specific molecular interactions that orchestrate and modulate regulation. The integrated contributions of genomics, proteomics and bioinformatics hold the promise of identifying functionally critical molecular interactions associated with complex biological networks, such as the interactions between regulatory elements/proteins and DNA that lie at the heart of gene regulation. To understand the properties of a biological network, quantitative information on the processes governing the interactions of network components is required. Consequently, the rise of genomics, proteomics, bioinformatics and systems biology, rather than hailing the end of experimental biophysics, has challenged these experimentalists to develop methods that can rapidly evaluate the physical properties of hundreds, even thousands, of macromolecules and their interactions. Spectroscopic and calorimetric methods traditionally used for such characterizations have been and continue to be applied primarily to isolated reactions between purified reagents. To date, high throughput methods generally identify only tightly bound complexes, while missing lower affinity, transient interactions that may well be biologically relevant. Current high throughput methods also fail to account for the broad range of physicochemical variables contributing to the modulation of macromolecule interactions. In a recent report, Maerkl and Quake1 present a novel high throughput microfluidics approach, called mechanically induced trapping of molecular interactions (MITOMI). Their method augments traditional high-throughput microarray technology with microfluidic “button” arrays to physically entrap low affinity complexes. The reported detection limit for transcription factor-DNA binding is Kd ≈ 18 μM, with a global measurement error of 19%. Microfluidic devices are increasingly used in biological experimentation due to their potential for miniaturization and automation.2-6 A critical deficiency of current understanding in systems biology is a detailed thermodynamic description of the sequence specific recognition of DNA regulatory sites by transcription factors. Maerkl and Quake apply the MITOMI method to determine relative binding affinities as a function of DNA sequence of four eukaryotic basic helix-loop-helix (bHLH) transcription factors (TF); namely, two human TFs—A and B isoforms of MAX, and two yeast TFs—Pho4p and Cbflp. The bHLH transcription factors bind the enhancer box (E-box) consensus sequence 5′-CANNTG-3′, where N denotes any nucleotide. A 24 × 100 array of Cy5-labeled DNA is spotted onto an epoxy coated glass surface, with the spots varying by nucleotide sequence within the E-box motif and the amount of DNA. A microfluidic device is aligned with the DNA array and bonded to the surface. Each unit of the microfluidic device comprises a chamber containing one of the DNA spots and an adjoining detection chamber. The microfluidic device is regulated through micromechanical valves that control the flow of materials through the units, as well as communication between the chambers. The surface of the detection chamber is coated with penta-histidine specific antibodies. His-tagged TFs are synthesized in situ and bind to the surface attached antibodies. DNA is introduced into the detection chamber and equilibrated with the surface bound TFs. The “button,” a deflectable membrane, is expanded to physically sequester the surface bound materials from the unbound material in the surrounding solution. The unbound DNA is washed away, while the bound TFs are quantified fluorometrically through the remaining bound labeled DNA. Based on the amounts of bound DNA detected, dissociation constants are calculated for each E-box variant and binding energy landscapes are constructed. The binding energy landscapes measured for all four TFs reproduce the expected optimal binding sequence 5′-CACGTG-3′. Interestingly, the human TFs showed more fine structure in the binding energy landscapes, indicating interactions with previously unrecognized low affinity binding sites, including consensus neighbors and sites with a one base spacer between the half sites. In addition to the identification of preferred binding sites, binding energy landscapes can be interpreted to answer a variety of biologically significant questions. Because the consensus sequence for the two yeast TFs is identical, their different functions indicate that additional interactions must distinguish the binding of the two molecules. Indeed, examination of sequences flanking the E-box consensus sequence indicates that the two yeast TFs recognize expanded consensus motifs. Maerkl and Quake further used the binding energy landscapes of the two yeast TFs to identify genes that each is likely to regulate. Combining a probabilistic model with the binding energy landscapes, they found that these genes are enriched for activities associated with phosphate metabolism, ionic homoeostasis, and vacuoles (Pho4p), and chromosome structure and budding (Cbflp). These activity patterns overlap partially with those assessed by other techniques and represent a concrete prediction of biological behavior. It has been assumed that the binding specificity of bHLH TFs is imparted solely by the basic region. The binding energy landscapes determined for chimeras constructed from the human MAX isoform B backbone and the basic regions of the two yeast TFs indicate that indeed the base region determines most if not all of the binding specificity. Bioinformatic methods for TF binding site discovery usually include the assumption that the energetic contribution to TF binding of a nucleotide base is independent of its neighbors. By using MITOMI to examine the binding of MAX isoforms A and B to all 256 permutations of one E-box half site plus a flanking base, Maerkl and Quake show that relative binding affinities predicted using this additivity assumption systematically fail to predict low affinity binding motifs. Thus, traditional bioinformatic methods may fail to identify a large number of potential binding sites. It is important to appreciate that the binding events governing gene regulation are controlled by physicochemical processes fully describable in terms of kinetics and thermodynamics. The binding energy landscapes determined by the MITOMI method begin to provide the requisite data needed to assess the thermodynamic origins of TF function. Due to a lack of data for comparison, it is difficult to assess whether the MITOMI method reliably quantifies the dissociation constants; however, for many applications, relative values are sufficient. A complete understanding of the thermodynamic and kinetic underpinnings of gene regulation networks will require discovery of additional new high throughput methods. By combining array-based assays with microfluidics and a novel means of preserving low affinity interactions, the MITOMI method described by Maerkl and Quake represents a significant step forward.