 |
Mechanistic Principles of Signal Processing in Cellular Protein Networks
Summary: Rama Ranganathan is interested in understanding the structural principles of function in cellular signaling systems and how these systems are built through the process of evolution.
Cells collect and process information about their external environment through the action of elaborate networks of proteins that we call the signal transduction machinery. This machinery converts the dynamics of extracellular signals into fluctuations in protein activities and concentrations of small molecules—an intracellular code that is then read out by other cellular mechanisms to make appropriate responses to input stimuli. The signal-processing capability of the transduction machinery is prodigious; weak input signals embedded in noise can be selectively amplified and filtered to provide efficient and reliable sensing of physiologically relevant stimuli, and many layers of positive- and negative-feedback control provide for complex nonlinear characteristics and protection against output saturation. Thus, for example, sensory neurons in our eye that mediate vision can reliably generate substantial and fast electrical responses to the absorption of single photons under dark-adapted conditions, but can also adjust the gain to continue operating in bright daylight conditions without response saturation.
A central objective in our work is to define the mechanistic principles in the evolutionary design of proteins and protein networks that encode these and other complex signal-processing capabilities. Two main problems limit our understanding of these systems at the current time. First, what are the structural and dynamic properties of proteins that provide for efficient molecular function, and how are they built through the process of evolution? Individual protein domains display specific binding, catalysis, signal transmission, and allosteric regulation; these properties constitute the elementary building blocks of larger multiprotein signal-processing devices in cells. Second, at the macromolecular level, what are the principles of organizing the signaling machinery in cells? Much evidence now supports the view that signaling is a highly structured process mediated by preformed protein complexes that operate in defined local compartments. How exactly does this organization contribute to signal processing? These questions encompass a broad set of studies at both atomic and cellular spatial scales but are unified in the goal of defining and understanding the core devices that comprise signal-processing modules in cells.
Mapping Energy Transduction in Proteins
Information flow and processing in cells occurs through the stimulus-dependent regulation of intermolecular interactions and control of enzymatic activities. For example, G protein–coupled receptors (GPCRs) bind ligands at an externally accessible site, but this localized structural (and energetic) event propagates through the tertiary structure to arrive as conformational change at cytoplasmic domains that mediate interaction with heterotrimeric G proteins. Similarly, binding of activated receptors to one site on the α subunit of the G protein trimer leads to changes at distant sites to trigger catalysis of nucleotide exchange, unbinding of Gβγ subunits, and binding of effector molecules. At the atomic level, these processes imply mechanisms for the orderly transfer of energy within and between proteins; thus, a perturbation at one site (e.g., ligand binding or covalent modification) is “felt� reliably at distant functional sites. Obtaining a general understanding of these mechanisms is important for explaining protein function, but is also a step in defining the “design principles� in evolution that have produced these marvelous devices. However, the problem is complex; amino acid residues in proteins often make unequal and cooperative contributions to function, and these contributions are not obvious in atomic structures.
To attempt a new approach to this well-established problem, we developed a statistical method (the statistical coupling analysis, or SCA) for globally estimating the energetic interactions between amino acid residues in a protein. This method, which takes advantage of evolution as a large-scale experiment in mutation, is based on the simple proposition that the energetic coupling of two residues in a protein (whether for structural or functional reasons) should force the correlated evolution of those two sites. In principle, this coupling might be exposed through calculating the statistical covariation of two positions in a large and diverse multiple sequence alignment of a protein family. Initial tests of this method for one active-site residue in a small protein interaction module known as the PDZ domain showed three important findings: (1) most residues in the PDZ domain do not coevolve with the selected active site residue, (2) the few residues that do coevolve form a physically connected network of van der Waals interactions within the protein core that connect the active-site positions with specific sites on the backside of the domain (Figure 1), and (3) the amino acid interactions predicted by the statistical method showed good correlation with experimental measurements of coupling through mutagenesis. This work led to the proposal that proteins may contain sparse but conserved networks of energetic coupling that represent the physical basis for information flow between distant sites.
If pathways of evolutionarily coupled residues are a general feature underlying function in proteins, then we should find such pathways in well-established model systems with strong experimental evidence for the underlying structural mechanisms. We tested this hypothesis in several cases: first in the GPCRs, the chymotrypsin class of serine proteases, and hemoglobin, families in which a large body of prior work has led to a mapping of functional mechanism, and then in the guanine-nucleotide-binding (G) proteins, and the ligand-binding domains (LBDs) of the heterotrimeric nuclear hormone receptors, in which we tested the predictions of the statistical analysis. The data show that a network of coevolving residues occurs in each protein family (Figure 2) that is consistent with the functional mechanism. Despite large differences in structure, mechanisms, and biological role, all of the proteins studied contain a surprisingly simple architecture of amino acid interactions: most residues act as if nearly evolutionarily independent, and a few (~10–15 percent) comprise an interconnected network of coevolving residues that link distantly positioned functional surfaces. Such an architecture is structurally nonintuitive but is consistent with empirical findings that proteins are tolerant to mutagenesis nearly everywhere while being exquisitely sensitive at a few sites. (Grants from the Burroughs Wellcome Fund and the Welch Foundation provided partial support for this work.)
Evolution's Rules for Building Proteins
The classic work of Christian Anfinsen showed that for many proteins, the amino acid sequence contains sufficient information to specify the native structure. A calculation of all the combinatorial ways that amino acid residues could interact shows, however, that the information density of a protein sequence could be enormous. Instead, the relative simplicity in the energetics implied by the SCA suggests that all the information required for specifying the fold and characteristic function of a protein family may be sufficiently encoded in just the small set of amino acid interactions revealed by the coevolution analysis. If so, the proof lies in building artificial members of a protein family, using only information extracted by the SCA. We developed a computational method for creation of artificial amino acid sequences according to the statistical rules revealed by the SCA and tested large libraries of designed sequences in the laboratory. We find that for the WW domain, a small β-structured protein fold, the computational method produces artificial sequences that efficiently fold into the characteristic WW domain structure and, remarkably, display the characteristic binding specificity of WW domains for polyproline-containing peptides. In contrast, randomly designed sequences, or even sequences preserving the conservation pattern of the WW domain but excluding the coevolution of residues, failed to show folding. Our current work is aimed at testing the sufficiency of the coevolution rules for the computational design of other functional proteins and at understanding how these results influence models of the evolution of protein families. (Grants from the Welch Foundation and the Mallinckrodt Scholars Program provided partial support for this work.)
Structure and Dynamics of Multiprotein Complexes in Signaling
To understand how protein macromolecular complexes work to facilitate signaling, we have carried out structural and functional analyses of InaD, a five-PDZ-domain scaffolding protein that organizes several components of the Drosophila visual signaling cascade into a single molecular unit at the light-sensitive membrane. PDZ domains are small protein interaction modules whose biological activity is typically thought to constitute binding and localization of target proteins to specific cellular sites. High-resolution crystal structures of one InaD PDZ domain (PDZ5) showed an unexpected finding: this domain is able to switch between two distinct conformations that differ in the geometry of the active site. In addition, the structure revealed a plausible hypothesis about the switching mechanism. To test this, we made transgenic flies expressing mutant forms of InaD that are defective in conformational switching of PDZ5. Electrophysiological studies of these mutant photoreceptor cells show that while many aspects of visual signaling are entirely normal, these cells are severely defective in their ability to adapt to strong light stimuli. The data support a model in which the InaD PDZ5 domain is structurally plastic under physiological conditions, and that this plasticity is important for visual signaling. Further studies are under way to understand how conformational switching of PDZ5 is triggered in vivo, and to test the model that InaD is a dynamic scaffold in vivo, regulating its composition as a function of the history of signaling activity. (Grants from the Burroughs Wellcome Fund provided partial support for this work.)
Last updated: January 12, 2006
|
|
|
|
