Agathe Urvoas

Artificial proteins from combinatorial approaches

How do we create new artificial proteins? In this review, we present a range of experimental approaches based on combinatorial and directed evolution methods used to explore sequence space and recreate structured or active proteins. These approaches can help to understand constraints of natural evolution and can lead to new useful proteins. Strategies such as binary patterning or modular assembly can efficiently speed structural and functional innovation. Many natural protein architectures are symmetric or repeated and presumably have emerged by coalescence of simpler fragments. This process can be experimentally reproduced; a range of artificial proteins obtained from idealized fragments has recently been described and some of these have already found direct applications.

Amphipol-mediated screening of molecular orthoses specific for membrane protein targets.

Specific, tight-binding protein partners are valuable helpers to facilitate membrane protein (MP) crystallization, because they can i) stabilize the protein, ii) reduce its conformational heterogeneity, and iii) increase the polar surface from which well-ordered crystals can grow. The design and production of a new family of synthetic scaffolds (dubbed αReps, for “artificial alpha repeat protein”) have been recently described. The stabilization and immobilization of MPs in a functional state are an absolute prerequisite for the screening of binders that recognize specifically their native conformation. We present here a general procedure for the selection of αReps specific of any MP. It relies on the use of biotinylated amphipols, which act as a universal “Velcro” to stabilize, and immobilize MP targets onto streptavidin-coated solid supports, thus doing away with the need to tag the protein itself.

Selection of Specific Protein Binders for Pre-Defined Targets from an Optimized Library of Artificial Helicoidal Repeat Proteins (alphaRep)

We previously designed a new family of artificial proteins named αRep based on a subgroup of thermostable helicoidal HEAT-like repeats. We have now assembled a large optimized αRep library. In this library, the side chains at each variable position are not fully randomized but instead encoded by a distribution of codons based on the natural frequency of side chains of the natural repeats family. The library construction is based on a polymerization of micro-genes and therefore results in a distribution of proteins with a variable number of repeats. We improved the library construction process using a “filtration” procedure to retain only fully coding modules that were recombined to recreate sequence diversity. The final library named Lib2.1 contains 1.7×109 independent clones. Here, we used phage display to select, from the previously described library or from the new library, new specific αRep proteins binding to four different non-related predefined protein targets. Specific binders were selected in each case. The results show that binders with various sizes are selected including relatively long sequences, with up to 7 repeats. ITC-measured affinities vary with Kd values ranging from micromolar to nanomolar ranges. The formation of complexes is associated with a significant thermal stabilization of the bound target protein. The crystal structures of two complexes between αRep and their cognate targets were solved and show that the new interfaces are established by the variable surfaces of the repeated modules, as well by the variable N-cap residues. These results suggest that αRep library is a new and versatile source of tight and specific binding proteins with favorable biophysical properties.

Specific GFP-binding artificial proteins (αRep): a new tool for in vitro to live cell applications

Artificial proteins, named αRep, binding tightly and specifically to EGFP are described. The structures of αRep–EGFP complexes explain how αRep recognize their cognate partner. Specific αRep can be used for biochemical or live cells experiments.

Disulfide Bond Substitution by Directed Evolution in an Engineered Binding Protein

Breaking ties: The antitumour protein, neocarzinostatin (NCS), is one of the few drug‐carrying proteins used in human therapeutics. However, the presence of disulfide bonds limits this proteins potential development for many applications. This study describes a generic directed‐evolution approach starting from NCS‐3.24 (shown in the figure complexed with two testosterone molecules) to engineer stable disulfide‐free NCS variants suitable for a variety of purposes, including intracellular applications.

Artificial Metalloenzymes with the Neocarzinostatin Scaffold: Toward a Biocatalyst for the Diels-Alder Reaction.

A copper(II) cofactor coupled to a testosterone anchor, copper(II)‐(5‐(Piperazin‐1‐yl)‐1,10‐phenanthroline)testosterone‐17‐hemisuccinamide (10) was synthesized and associated with a neocarzinostatin variant, NCS‐3.24 (KD=3 μm), thus generating a new artificial metalloenzyme by following a “Trojan horse” strategy. Interestingly, the artificial enzyme was able to efficiently catalyze the Diels–Alder cyclization reaction of cyclopentadiene (1) with 2‐azachalcone (2). In comparison with what was observed with cofactor 10 alone, the artificial enzymes favored formation of the exo products (endo/exo ratios of 84:16 and 62:38, respectively, after 12 h). Molecular modeling studies assigned the synergy between the copper complex and the testosterone (KD=13 μm) moieties in the binding of 10 to good van der Waals complementarity. Moreover, by pushing the modeling exercise to its limits, we hypothesize on the molecular grounds that are responsible for the observed selectivity.

Structural and functional analysis of the fibronectin-binding protein FNE from Streptococcus equi spp. equi

Streptococcus equi is a horse pathogen belonging to Lancefield group C. Infection by S. equi ssp. equi causes strangles, a serious and highly contagious disease of the upper respiratory tract. S. equi ssp. equi secretes a fibronectin (Fn)‐binding protein, FNE, that does not contain cell wall‐anchoring motifs. FNE binds to the gelatin‐binding domain (GBD) of Fn, composed of the motifs 6FI12FII789FI. FNE lacks the canonical Fn‐binding peptide repeats observed in many microbial surface components recognizing adhesive matrix molecules. We found that the interaction between FNE and the human GBD is mediated by the binding of the disordered C‐terminal region (residues 208–262) of FNE to the 789FI GBD subfragment. The crystal structure of FNE showed that it is similar to the minor pilus protein Spy0125 of Streptococcus pyogenes, found at the end of pilus polymers and responsible for adhesion. FNE and Spy0125 both have a superimposable internal thioester bond between highly conserved Cys and Gln residues. Small‐angle X‐ray scattering of the FNE–789FI complex provided a model that aligns the C‐terminal peptide of FNE with the E‐strands of the FI domains, adopting the β‐zipper extension model observed in previous structures of microbial surface components recognizing adhesive matrix molecule adhesion peptides bound to FI domains.

The unexpected structure of the designed protein Octarellin V.1 forms a challenge for protein structure prediction tools.

Despite impressive successes in protein design, designing a well-folded protein of more 100 amino acids de novo remains a formidable challenge. Exploiting the promising biophysical features of the artificial protein Octarellin V, we improved this protein by directed evolution, thus creating a more stable and soluble protein: Octarellin V.1. Next, we obtained crystals of Octarellin V.1 in complex with crystallization chaperons and determined the tertiary structure. The experimental structure of Octarellin V.1 differs from its in silico design: the (αβα) sandwich architecture bears some resemblance to a Rossman-like fold instead of the intended TIM-barrel fold. This surprising result gave us a unique and attractive opportunity to test the state of the art in protein structure prediction, using this artificial protein free of any natural selection. We tested 13 automated webservers for protein structure prediction and found none of them to predict the actual structure. More than 50% of them predicted a TIM-barrel fold, i.e. the structure we set out to design more than 10years ago. In addition, local software runs that are human operated can sample a structure similar to the experimental one but fail in selecting it, suggesting that the scoring and ranking functions should be improved. We propose that artificial proteins could be used as tools to test the accuracy of protein structure prediction algorithms, because their lack of evolutionary pressure and unique sequences features.

The αRep artificial repeat protein scaffold: a new tool for crystallization and live cell applications.

We have designed a new family of artificial proteins, named αRep, based on HEAT (acronym for Huntingtin, elongation factor 3 (EF3), protein pphosphatase 2A (PP2A), yeast kinase Tor1) repeat proteins containing an α-helical repeated motif. The sequence of the repeated motifs, first identified in a thermostable archae protein was optimized using a consensus design strategy and used for the construction of a library of artificial proteins. All proteins from this library share the same general fold but differ both in the number of repeats and in five highly randomized amino acid positions within each repeat. The randomized side chains altogether provide a hypervariable surface on αRep variants. Sequences from this library are efficiently expressed as soluble, folded and very stable proteins. αRep binders with high affinity for various protein targets were selected by phage display. Low micromolar to nanomolar dissociation constants between partners were measured and the structures of several complexes (specific αRep/protein target) were solved by X-ray crystallography. Using GFP as a model target, it was demonstrated that αReps can be used as bait in pull-down experiments. αReps can be expressed in eukaryotic cells and specifically interact with their target addressed to different cell compartments.