- Research Overview
- Chemical Biology and Medicinal Chemistry
- Chemical Biology and Medicinal Chemistry Overview
- Discovering Enzyme Substrates and Functions
- Discovering Protein Ligands to Probe and Alter Function
- Discovering Enzyme Activators
- Analyzing Mechanisms of Drug Resistance via Chemical Biology
- Analyzing Enzyme Conformational Dynamics, Substrate Binding, and Catalysis
- Effective Drug Targeting of Pathogens via Medicinal Chemistry
- Computational Chemistry and Biology
- Computational Chemistry and Biology Overview
- Modeling protein regulation via allostery and post-translational modifications
- Visualizing and integrating bioinformatics and biomolecular data
- Modeling membrane permeation to optimize pharmacokinetics
- Determining enzyme function by predicting substrate specificity
- Physical Biology
- Protein and Cellular Engineering
- Protein and Cellular Engineering Overview
- Monitoring enzyme activity and disease biomarkers
- Generating human proteome antibodies via phage display and directed evolution
- Globally analyzing and dissecting apoptosis
- Proximity tagging of protein-protein interactions
- Investigating cellular interactions in tissues
- Creating fluorescent probes targeting the genome and key bio-pathways
- De novo design of catalytic and membrane proteins
- Probing and modulating membrane proteins
Modeling protein regulation via allostery and post-translational modifications
Examples of our research and methods include
Determining how post-translational modifications change protein behavior
Post-translational modifications (PTMs) encompass a wide variety of changes to proteins after they are translated from genetic codes into chains of amino acids (residues) linked by peptide bonds. Such changes include the addition of chemical groups at specific locations (typically by enzymes which may themselves be modified in sequential biological pathways) that can cause changes in function. This allows both rapid responses to stimuli and the regulation of key processes. Such regulation can go awry in diseases, such as the over-activated PTM-mediated cell growth pathways in cancers.
But the understanding of how particular modifications alter protein function lags far behind. Physical characterization is possible via time-consuming higher resolution technologies (e.g., x-ray crystallography, NMR spectroscopy) but it can be challenging to obtain sufficient amounts of purified proteins complete with a particular PTM for study.
Computer simulations can thus be a faster and vital way to bridge the growing gap between the discovery of PTMs and the determination of their regulatory mechanisms and effects. Specifically, such simulations can help parse how the addition of chemical groups at specific locations alters the energy landscape of a protein, re-arranging the atomic structure to stabilize new shapes that yield changed function.
Such simulations depict atoms as charged spheres of different radii connected by springs representing chemical bonds. The approach calculates the total energy of the protein in a given state, accounting for the key physical forces, thus making it possible to predict how the molecular energy landscape is altered by a binding event. Molecular dynamics simulations track the atomic interactions over time, using Newton’s laws of motion. Given the complexity of applying mathematical functions to complex systems of thousands of interacting atoms in multiple dimensions—plus accounting for the effects of surrounding solvent—even simulations run on supercomputers can currently only capture fractions of seconds of molecular motion for analysis.
It is estimated that up to 30 percent of all proteins are phosphorylated—have one or more phosphate groups (PO4) added to specific amino acids by kinase enzymes—often multiple times, with an estimated half-million phosphorylation sites in the human proteome and tens of thousands of specific sites currently listed in the database Phospho.ELM.
This ubiquitous, rapid, reversible PTM (phosphatases remove phosphates) routinely acts as a switch that can turn proteins’ activity on and off in response to stimuli or tune their activity more subtly. It plays a regulatory role in cellular signaling, cytoskeletal construction, metabolism, and the cell cycle (replication). It goes awry in numerous diseases, including many forms of cancer.
Department scientists use computation to simulate, analyze, and predict changes in conformation and protein motion wrought by phosphorylation at specific sites in a protein.
How does phosphorylation change a protein's structure?
In an early example of simulating protein phosphorylation in silico, researchers here demonstrated that it was possible to computationally predict local phosphorylation-driven conformational changes in secondary molecular structures such as loops and helices with near-atomic accuracy, given a pre-determined structure and modification site.
A molecular mechanics method was tested for accuracy using proteins which had had both phosphorylated and non-phosphorylated structures determined by x-ray crystallography. In a case study, department scientists predicted and analyzed changes due to phosphorylation that lead to the activation of a cyclin-dependent kinase, CDK2, that plays a key role in cell division and replication.
Specifically, researchers here adapted loop prediction algorithms previously developed for homology modeling to predict changes to the CDK2 activation loop upon phosphorylation of a constituent threonine residue. (Most kinases are activated by phosphorylation of such loop structures).
This approach first hierarchically sampled the potential backbone dihedral angles, reducing the need to model from scratch all potential conformations of the 11-residue loop, with and without the added phosphate, by eliminating modeled structures with sterically impossible overlapping of atomic radii. Possible loop conformations then had residue side chains, and those in the adjacent structure, optimized and energy minimized.
Indeed, such modeling could be used to generate mechanistic hypotheses about phosphorylation regulation for experimental testing (e.g., altering key residues via site-directed mutagenesis). For example, this early case study indicated that phosphorylation not only changed activation loop structure (a modified threonine residue in the loop moves to interact with positively charged arginine side chains) but also revealed that it rearranges extended side-chain hydrogen bonding beyond the loop.
Analyzing activation of a cytoskeletal complex
Department researchers and colleagues used molecular dynamics simulations to detail how phosphorylation of actin-related protein 2 (Arp2) induces conformational change in the highly regulated Arp2/3 complex, thus allowing its activation by binding yet other proteins, including nucleation-promoting factors (NPFs).
Arp2 assembles with six other protein subunits to form the complex, which plays a central role in forming actin filaments that are assembled into cellular structures used for many purposes, including cell movement (motility). Dysregulation of the complex’s activity been linked to cancer metastasis.
While X-ray crystallography of Arp2/3 in unbound and bound structures suggested that large structural changes in the complex are required prior to complete activation, computational modeling helped to show how and why those changes come about.
Given PTM locations determined by mass spectrometry, molecular dynamics simulations modeled the addition of a -2 charged phosphate group to two uncharged threonine amino acids. Despite simulation timescale limitations (30 nanoseconds), there were significantly larger structural changes in the phosphorylated Arp2 versus a control simulation of unphosphorylated Arp2. Moreover, the simulations suggest that the electrostatic changes destabilize a network of salt bridges at the interface of several sub-units, causing a reorientation of Arp2 that permits the complex’s activation.
In addition, researchers created an in silico mutant form of the Arp2/3 complex to test whether changing positively charged arginine residues that interact with the phosphorylated Arp2 threonines to alanines would block the conformational changes. Instead, the simulations predicted that the mutant form would increase activity, akin to the effect of phosphorylation. Ensuing in vitro experiments confirmed that the mutant form was substantially active even in the absence NPFs.
Assessing hydrogen bond strength of phosophorylated side chains
Phosphorylated residues are noted for accepting H-bonds through their phosphate oxygens, often with positively charged donor side chains (arginine, lysine) and backbone amide groups. These hydrogen bonds stabilize PTM-altered conformations that can drive changes in function.
Using small molecule analogs for phosphorylated side chain acceptors (e.g., methyl phosphate for phospho-serine and phospho-threonine) and donors (e.g., butyl ammonium for lysine side chains), researchers quantified the relative strengths of hydrogen bonds (measured in potential means of force) for common partners, charge states, and geometries.
Department scientists used multiple types of simulation to analyze the relative strengths of hydrogen bonds and salt bridges involving phosphorylated side chains. These included molecular dynamics simulations with explicit solvent and multiple bonding geometries (hydrogen bond energies being sensitive to the solvent environment, in addition to the identity, proximity, and orientation of participating side chains); and quantum mechanics calculations that sought to account for asymmetric distributions of electrons (partial charges) in phosphorylated residues, and shifts in electrons wrought by phosphates’ strong electric field which impact the polarities of neighboring atoms.
Specifically, the researchers found arginine capable of forming very strong hydrogen bonding interactions with phospho-serine, possibly the strongest hydrogen bonding interaction commonly found in proteins. The simulations also helped explain why a frequently used in vitro model for protein activation via phosphorylation which substitutes residues with negatively charged side chains (aspartic and glutamic acid) for phosphorylated ones is sometimes ineffective (i.e., phosopho-serines form stronger salt bridge with charged residue donors in all simulated orientations).
The simulations also had implications for designing inhibitors of SH2 protein domains, which bind with phosphorylated tyrosine residues on other proteins to pass signals across membranes related to cell growth, making them a target for cancer drugs.
pH regulation of proteins
The cytoplasm of most cells remains very close to neutral pH, but there are small fluctuations over time in many cells, and there are differences in pH between normal and cancer cells. Department scientists along with UCSF colleagues are exploring an emergent view that small shifts in intracellular pH may regulate a number of vital cell processes, including proliferation and movement.
Since pH is a measure of hydrogen ion (H+, also called a proton) concentration, the researchers propose that its intracellular dynamics can be seen as a posttranslational modification that reversibly adds and removes protons within physiological pH ranges. The addition or removal of a proton alters side chain charges, and can change a hydrogen bond acceptor to a potential donor. As with other modifications like phosphorylation, these changes can alter the structure and function of proteins.
A minority of protein sites are candidates for protonation at the normal intracellular pH range, including those with appropriately positioned histidine side chains. Nonetheless, a variety of proteins, dubbed pH sensors, can be altered by the modification. As with other PTMs, computational methods such as molecular dynamics simulation can predict relatively rapidly and inexpensively how pH-related protonation affects a pH sensor’s structure and conformational dynamics, thus suggesting functional changes that can then be experimentally investigated.
Determining a mechanism for pH regulation of cell migration
Cell migration in tissues is routinely required for wound healing and immune response but it goes awry in cancer cell invasion and metastasis.
In order to move, cells require cytoskeletal elements to extend the cell in a particular direction. To gain traction and pull itself forward, the cell relies on macromolecular assemblies called focal adhesions which link those leading-edge cytoskeletal elements to transmembrane integrin proteins that bind to extracellular surfaces. The intracellular domains of integrins bind to adapter proteins such as talin that, in turn, bind to cytoskeletal filamentous actin. Focal adhesions are assembled and disassembled as the cell moves forward.
It is known that increased intracellular pH (alkalinity) promotes cell migration and is also seen in a variety of metastatic cancers. Researchers here used molecular dynamics simulations to explore a mechanism underlying this, specifically by analyzing a module of talin that showed lower actin binding at higher pH. Computational simulations showed that when the module’s sole histidine was protonated (as under low pH conditions), it induced changes in the conformation and dynamics of a distant binding site.
This allosteric effect of the specific protonation was experimentally confirmed by synthesizing a mutant talin with the histidine replaced by phenylalanine. The mutant talin displayed actin-binding that was relatively pH-insensitive. Expressed in mobile cells, it increased the longevity of focal adhesions and decreased migration.
Discovering allosteric sites and mechanisms
Allostery offers a selective way to therapeutically modulate protein function via small molecule binding at an alternative site that affects active site conformation (or the distribution of active site conformations over time), and thus protein function, such as binding or catalysis. However, allosteric sites are often discovered only by chance and, as with PTMs, there is limited information about the specific atomic-level mechanisms by which binding at one site causes a change at a different location.
Department scientists have developed an unbiased approach for identifying intramolecular allosteric networks and discovering new allosteric sites. The researchers hypothesized that correlated motions and conformational changes observed in residues at different sites in a protein at equilibrium—but subject to small on-going changes in its energy landscape, such as collisions with solvent water molecules—would follow the same fluctuation pathways induced by a small molecule binding event at an allosteric site.
Their approach, dubbed Mutual Information (MutInf), quantifies significant correlations between the atomic-scale motions of residues at different sites during multiple brief (10 nanosecond) molecular dynamics simulations of a protein’s movements at equilibrium. Since such linked changes in motions at different sites may occur sequentially rather than concurrently, the approach’s calculations looked for correlations in the statistical distributions of the residue’s motions and conformations over time.
While the atomic components of molecules at equilibrium may oscillate around a native geometry, MutInf focuses on low-frequency motions such as gear-like twists of residue side chains around protein backbones. The approach correlates a sub-set of internal coordinates—certain backbone and side chain torsion angles which the researchers believed to be most relevant for discerning biologically important motions.
Proof of principle study: interleukin-2
Researchers here applied the MutInf approach to identify allosteric mechanisms in human interleukin-2 (IL-2), a key immune system protein. IL-2 is known to exhibit positive cooperativity—that is, it increases its active site affinity for ligands upon small molecule binding at an allosteric site nearly seven angstroms distant.
Since x-ray crystallography structures of IL-2 bound with small molecules did not reveal large changes at other locations, the scientists hypothesized that the allostery arose from correlated changes in motions and subtle shifts in conformation distributions of cooperative sites, rather than a major change in backbone conformation.
Simulations of IL-2 at equilibrium were analyzed for significant correlations between the twisting motions of residue side chains. Indeed, two residue clusters with the strongest correlated movements represented known cooperative binding sites, suggesting the approach could indeed be applied to identify unknown allosteric sites.
Additional simulations of the protein in complex with ligands (at allosteric or active site) demonstrated the mechanism of cooperativity between the two sites: Interactions both via the protein’s hydrophobic core and also through a dynamic polar network of hydrogen bonding and electrostatic interactions. This mechanism ultimately leads to changes in the distribution of motions of two residues’ side chains at the active site that make it more accessible to ligand binding. This shift in positive cooperativity was qualitatively confirmed by virtually docking small molecules to simulation “snapshots.”