59th Annual Report on Research 2014

Monomer Redesign for Solubility

We selected an ambitious target for de novo crystal design. Namely, we aimed to redesign a fragment of a Methane Monooxygenase, with the ultimate goal being a solid state, crystalline catalyst for conversion of methane to methanol. Unfortunately, the starting fragment, originally described by Rosenszweig and coworkers is difficult to express and highly unstable. We were able to express the codon optimized wild-type fragment in E. coli, and to identify it on denaturing electrophoresis gels.       Prior to designing a crystalline assembly thereof, our first goal was to redesign the fragment for increased stability and solubility. Since the fragment was cut out of a large domain in the original methanotropic organism, it exposed hydrophobic surface area. Our design strategy emulated the principles that were proposed in the original research proposal. Namely, we attempted to optimize the electrostatics. We had three variant genes synthesized. Two were superpositive, with a predicted net charge of 28.6 and 27.5 (at pH 8.0). A third was supernegative, with a predicted net charge of -12.5 (at pH 8.0). The resulting electrostatic potential (Fig. 1) as assessed using Poisson Boltzmann calculations was intended to improve solubility and to decrease aggregation.

            Unfortunately, repeated efforts to isolate these variants proved unsuccessful. Expression appears to be the key challenge. Despite the small size (152 residues, including a 22-residue leader sequence) and enhanced hydrophilicity of the sequences, we were not able to isolate a significant amount of the variants via electrophoresis.

Rational Crystal Interface Redesign

In parallel with the ambitious project, we attempted to fulfill the goals of the project via crystal redesign. Rather than attempting to validate computational design principles for crystal assembly, we took a model protein crystal (PDB entry 2fgs) and used computational modeling and design methods to identify opportunities to covalent bonds across the interface. The goal was to convert the weak, non-covalent, assembly of trillions of monomers comprising such crystals into a robust, covalent crystal.       Perhaps the most natural possibility is to encode crosslinks using disulfide bonds. Previous studies have established that disulfide bonds can form spontaneously under the oxidizing conditions typical for protein crystals, though disulfide formation kinetics have varied. Bustamante and coworkers introduced cysteine residues at crystal contacts and observed moderate polymerization after exposure to pure oxygen gas for weeks. In contrast, Bell and coworkers observed extensive polymerization in crystallo upon exposure to atmosphere for days.       We designed an algorithm for scanning interfacial sites for potential opportunities to introduce covalent bonds. First, we search for possible intermolecular disulfide bonds of geometric quality comparable to the disulfide bonds found in a survey of high-resolution protein structures (<2) from the PDB. While disulfide design is not novel, our program has additional capabilities. First, we increase the number of viable disulfides by generating local backbone variations using a kinematic approach. Second, we also search for sites where cysteines could be placed that would support crosslinking using chemical crosslinkers such as bismaleimidoethane (BMOE, Fig. 2a). We plan to publish the code necessary for this rational, chemical crosslinking, and to make the code available as part of our open-source protein modeling software platform, SHARPEN.       We grew crystals for two designs intended to yield intermolecular disulfides: 2fgs-G127C, and 2fgs-I42C-I173C. Standard crystallization conditions were augmented with dithiothreitol (DTT), a reducing agent. To quantify the crosslink yield we dissolve crystals and perform denaturing gel electrophoresis. Preliminary results are encouraging: only the expected dimer band was present when a 3-week-old 2fgs-I42C-I173C crystal was dissolved and run on a gel. We plan to publish this result once we obtain a crystal structure that validates the formation of the intended disulfide bonds.       Our program also identifies sites suitable for the formation of dityrosine crosslinks (Fig. 2b). Dityrosine crosslinks are used in nature to impart unusual structural stability to protein-based materials, and can be induced in the laboratory using a variety of methods. We have very recently grown crystals for three designs intended to yield intermolecular dityrosine bonds: 2fgs-G193Y, 2fgs-S194Y, 2fgs-S176Y. We were surprised to discover that wild-type crystals also became quite stable during certain oxidation tests. Publication of the successful oxidative crosslinking results, has been slowed by the difficulty of definitively establishing that the observed crystal stability is indeed due to dityrosine formation. We have recently grown control crystals, 2fgs-Y99F, in which no dityrosine crosslinks should be possible. We are currently attempting to identify oxidative crosslinking protocols that are selective for dityrosine crosslinks.

Optimizing Adventitious Crosslinking

To better understand the constraints on rational, chemical crosslinking of protein-protein interfaces within crystals, we performed a comprehensive suite of hen egg white lysozyme crosslinking experiments. While we were able to stabilize the crystals using a variety of chemical crosslinking agents, and to obtain numerous high resolution structures, we were not able to observe the resulting crosslinks in the electron density map. We concluded that existing crosslinking protocols were likely crosslinking only the exterior layers of the crystal.

Efficient Crystal Search Algorithm in Hexagonal Space Groups

The majority of the algorithm development effort was finished in the first year of the grant. However, there were some barriers to completion. First, the underlying protein-protein docking scoring parameters needed to be optimized. Second, it was not clear how to extend the approach to crystal forms of higher symmetry (e.g. hexagonal forms). We have recently optimized the docking parameters to enable recapitulation of known high-density crystal forms, and extended the approach to hexagonal grids. We will be publishing the work, and making the crystal search code accessible. The same suite of tools will also be useful to thoroughly search the possible crystal forms for small molecule crystals or metal-organic frameworks.