Producing sufficient amounts or stable, properly folded protein is an essential prerequisite for protein crystallization, especially for challenging targets such as membrane proteins. However many proteins in their native form show limited stability and are refractory to expression in recombinant hosts. Over the last several years, ancestral sequence reconstruction has (ASR) emerged as a useful tool by which protein engineers and crystallographers can obtain highly robust forms of proteins of interest, which are often also expressed at relatively high levels in Escherichia coli. ASR involves inferring the ancestral state from an alignment of the sequences of extant forms of a given protein family, an evolutionary tree that represents their phylogeny, and an amino acid substitution model. We have developed a suite of software (GRASP: Graphical Representation of Ancestral Sequence Predictions) which enables the inference of ancestral protein sequences from alignments of up to ~ 10000 sequences using maximum likelihood, joint or marginal reconstruction methods. GRASP has been exemplified using several families of eukaryotic cytochrome P450 enzymes, membrane-bound, haemoprotein monooxygenases that have typically been challenging to express and crystallise. To date ASR has enabled expression of ancestral eukaryotic P450s at levels up to ~ 7 µM in E. coli cultures (~350 mg/L culture) leading to the successful crystallization of representative enzymes from several P450 subfamilies. While the resurrected ancestral proteins may not be identical to the extant proteins of principal interest, crystallization of ancestral homologues can provide insights into the structure of a protein family, including how to stabilize the protein fold. In addition, ancestors provide a robust and relevant template for structure function studies as well as protein engineering. When applied to previously uncharacterized sequences, ASR could enable the discovery of new folds and accelerate the functional and structural annotation of sequence-structure-function relationships in novel protein families.

Atomic crystal structure affects the electromagnetic and thermal properties of common matter. Similarly, the nanoscale structure controls the properties of higher length-scale metamaterials, for example nanoparticle superlattices and photonic crystals. We have investigated the self-assembly and characterization of binary solids that consist of crystalline arrays of 1) spherical viruses / other protein cages and 2) functional units [1]. The extremely well defined structure of protein cages (e.g. CCMV, TMV and ferritins) facilitates the construction of co-crystals with large domain sizes. The use of a second functional unit allows highly selective pre- or post-functionalization with different types of functional units, such as organic dyes [2,3], supramolecular hosts [4] and enzymes [5].

In the case of rod-like protein assemblies (e.g. tobacco mosaic virus), well-defined binary superlattice wires can be achieved [6]. The superlattice structures are explained by a cooperative assembly pathway that proceeds in a zipper-like manner after nucleation. Curiously, the formed superstructure shows right-handed helical twisting due to the right-handed structure of the virus. This leads to structure-dependent chiral plasmonic function of the material.

Our systematic approach identifies the key parameters for the assembly process (ionic strength, electrolyte valence, pH) and highlights the effect of the size and aspect ratio of the virus particles, which ultimately control the crystal structure and lattice constant. Protein-based mesoporous materials, nanoscale multicompartments and metamaterials are all applications that require such high degree of structural control.

[1] Kostiainen, M. A.; Hiekkataipale, P.; Laiho, A.; Lemieux, V.; Seitsonen, J.; Ruokolainen, J.; Ceci, P. (2013). Nat. Nanotech. 8, 52.

[2] Mikkilä, J., Anaya-Plaza, E., Liljeström, V., Caston, J. R., Torres, T.; De La Escosura, A. & Kostiainen, M. A. (2016). ACS Nano 10, 1565.

[3] Anaya-Plaza, E., Aljarilla, A., Beaune, G., Nonappa, Timonen, J. V. I., de la Escosura, A., Torres, T. & Kostiainen, M. A. (2019). Adv. Mat. 31, 1902582.

[4] Beyeh, N. K., Nonappa, Liljeström, V., Mikkilä, J., Korpi, A., Bochicchio, D., Pavan, G. M., Ikkala, O., Ras, R. H. A. & Kostiainen, M. A. (2018). ACS Nano 12, 8029.

[5] Liljeström, V., Mikkilä, J. & Kostiainen, M. A. (2014). Nature Commun. 5, 4445.

[6] Liljeström, V., Ora, A., Hassinen, J., Heilala, M., Hynninen, V., Joensuu, J., Nonappa, Rekola, H., Törmä, P., Ikkala, O., Ras, R. H. A. & Kostiainen, M. A. (2017). Nature Commun. 8, 671.

The functionality of a protein depends on its unique three-dimensional structure, which is a result of the folding process when the nascent polypeptide follows a funnel-like energy landscape to reach a global energy minimum. Computer-encoded algorithms are increasingly employed to stabilize native proteins for use in research and biotechnology applications [1]. Here, we reveal a unique example where the computational stabilization of a monomeric α/β-hydrolase fold enzyme (T_m = 73.5°C; ΔT_m > 23°C) affected the protein folding energy landscape. Introduction of eleven single-point stabilizing mutations based on force field calculations and evolutionary analysis yielded catalytically active domain-swapped intermediates trapped in local energy minima. Crystallographic structures revealed that these stabilizing mutations target cryptic hinge regions and newly introduced secondary interfaces, where they make extensive non-covalent interactions between the intertwined misfolded protomers [2]. The existence of domain-swapped dimers in a solution is further confirmed experimentally by data obtained from SAXS and crosslinking mass spectrometry. Unfolding experiments showed that the domain-swapped dimers can be irreversibly converted into native-like monomers, suggesting that the domain-swapping occurs exclusively in vivo [2]. Our findings uncovered hidden protein-folding consequences of computational protein design, which need to be taken into account when applying a rational stabilization to proteins of biological and pharmaceutical interest.

References

[1] Markova K., Chmelova K., Marques S. M., Carpentier P., Bednar D., Damborsky J., Marek M. (2020). Decoding the intricate network of molecular interactions of a hyperstable engineered biocatalyst. Chemical Science 11, 11162-11178.

[2] Markova K., Chmelova K., Marques S. M., Carpentier P., Bednar D., Damborsky J., Marek M. (2021). Computational protein stabilization can affect folding energy landscapes and lead to domain-swapped dimers. ChemRxiv. Preprint. https://doi.org/10.26434/chemrxiv.13634021.v1

L-asparaginases are used in the treatment of Acute Lymphoblastic Leukemia (ALL). However, the currently utilized bacterial-type enzymes cause several severe side effects. Therefore, there is an urgent need to develop a new generation of therapeutic L-asparaginases. Promising candidates can be found among plant-type L-asparaginases, which function as Ntn-hydrolases. Ntn‑hydrolases are produced as inactive precursors and develop enzymatic activity in an autoproteolytic maturation process. Unfortunetely, therapeutic use of wild-type (WT) plant-type enzymes is not possible due to their low substrate affinity (mM). Random mutagenesis is a protein engineering tool that can be used to enhance enzyme efficiency and study structure-function relationship. In this project, locally performed random mutagenesis was used to generate a library of mutants of plant-type L-asparaginase from E.coli (EcAIII). Several new variants of EcAIII were selected from the library and subjected to detailed structural (X-ray crystallography) and biophysical studies (nanoDSF, CD, activity/autoproteolytic tests).

The results of our studies revealed that autoproteolysis and enzymatic activity of EcAIII are unrelated events. Some variants of EcAIII retained the ability to autoprocess even in the absence of Arg207 (Fig. 1C), which is critical for catalysis and important for maintaining a H-bond network in the active site. Screening of thermal stability by nanoDSF showed that all analyzed unprocessed EcAIII variants had decreased T_m with respect to the WT enzyme. Thermal stability of variants cleaved into the mature aβ subunits varied, but some mutants with increased T_m were also found (Fig. 1A). Crystallization experiments proved that it was possible to obtain crystals of all variants cleaved into subunits. This was in stark contrast to the unprocessed mutants, which did not produce any crystals despite of extensive screening. This effect may be related to the presence of a highly disordered a-β linker in the uncleaved proteins. CD spectra showed that most of the unprocessed mutants are folded like the WT protein; however, some variants with significant changes in the CD signal were also found. The determined crystal structures (resolution 1.6-2.4 Å) showed that the active site of EcAIII is flexible enough to accept different amino acid substitutions; however, the type of substitution affects the H-bond pattern in the active site. Absence of Arg207 affects the overall conformation of the protein and leads to significant shifts of atomic positions in the entire enzyme molecule (Fig. 1B), as illustrated by the Cα rmsd value of 1.10 Å.

Work supported by National Science Centre (NCN, Poland) grants 2020/38/E/NZ1/00035 and 2019/03/X/NZ1/00584.

Pseudo symmetric, repeat proteins are favoured targets for computational protein design as they allow for the creation of larger domains with limited amino acids by exploiting their symmetric and repeating nature. One of the most common pseudo symmetric, repeat domains is the β-propellers fold. In addition, they fulfil many functions from sugar binding to enzymatic and protein-protein interaction mediation, thus increasing the potential applications of the designed proteins. Each propeller is built from 4-stranded anti-parallel β-sheets also known as a blade, repeated around a central axis. The number of blades differs from four to ten with seven and eight being the most common. The first successful computational protein design of a β-propeller was the 6-bladed Pizza protein¹. The RE3volutionary design method² makes use of ancestral sequence reconstruction and symmetry based template construction methods incorporated in Rosetta. Each blade of the pizza protein possess the same amino acid sequence. When two or three repeats of this sequence are expressed, they self-assemble into the 6-bladed domain.

The same design method was employed to design the eight or nine bladed Cake protein³. The protein consists repeating units of 42 amino acids, when eight repeats are expressed, the protein adopts the nine bladed fold. However, when nine repeats are expressed, the protein will adopt a nine bladed fold. This structural plasticity was unseen among β-propellers monomers. Its existence might explain the wide diversity of repeat numbers observed in β-propellers by allowing the change from even to odd numbers. Identical to the Pizza protein, smaller repeat fragments of Cake will self-assemble into either the eight-bladed protein or the nine-bladed protein. The structures of most these assemblies as well as the monomeric eight-and nine-bladed propellers were confirmed with X-ray diffraction experiments.

While the structural plasticity of the Cake protein is novel, we also wanted to create a protein that could only adopt the rare nine-bladed propeller fold. In order to achieve this a three-blade repeat (124 amino acids) was designed with a similar design strategy, with the idea the three-fold symmetry would prevent formation of eight bladed propellers. Two variants, Scone-E and Scone-R were created⁴. Crystallography revealed however that both designs adopted an eight-fold conformation. This failed design showcases that more research is needed to create a specific sequence for large β-propellers. In addition to this the Scone-E protein could only be crystallized upon addition of the polyoxometalate STA. This charged molecule interacts with multiple positively charged regions on the protein surface, neutralizing them. It can also bind multiple chains thus facilitating protein contacts, resulting in higher symmetric space groups.

Some of the designed proteins in this research behaved unexpectedly, thus illustrating the importance of accurate structure determination by X-ray diffraction to validate the designed proteins. In addition the design lessons on larger β-propellers could prove instrumental in the design of new functional proteins based on this common natural protein fold.

1. Computational design of a symmetrical β-propeller, Arnout R.D. Voet, Hiroki Noguchi, Christine Addy, David Simoncini, Daiki Terada, Satoru Unzai, Sam-Yong Park, Kam Y. J. Zhang, Jeremy R. H. Tame Proceedings of the National Academy of Sciences Oct 2014, 111 (42) 15102 15107; DOI: 10.1073/pnas.1412768111

2. Evolution-Inspired Computational Design of Symmetric Proteins, Arnout R. D. Voet, David Simoncini, Jeremy R. H. Tame, Kam Y. J. Zhang, Computational Protein Design, 2017, Volume 1529 ISBN : 978-1-4939-6635-6

3. Structural plasticity of a designer protein sheds light on β‐propeller protein evolution, Mylemans, B., Laier, I., Kamata, K., Akashi, S., Noguchi, H., Tame, J.R.H. and Voet, A.R.D. FEBS J, 2021, 288: 530-545. https://doi.org/10.1111/febs.15347

4. Crystal structures of Scone, pseudosymmetric folding of a symmetric designer protein, Bram Mylemans, Theo Killian, Laurens Vandebroek, Luc Van Meervelt, Jeremy R.H. Tame, Tatjana N. Parac-Vogt, Arnout R.D. Voet bioRxiv 2021.04.12.439409; doi:https://doi.org/10.1101/2021.04.12.439409

Fluorescent proteins (FPs) play an indispensable role in advanced imaging techniques. Such proteins are considered as “smart labels” allowing scientists to overcome the diffraction barrier of conventional light microscopy to visualize subcellular events. Since the first discovery of GFPs in the 1960s [1], numerous studies have been conducted to design new fluorophores not only covering the whole visible light from cyan to far-red region but also displaying improved photochemical performances. Furthermore, the FP technology gains remarkable achievements by developing successfully special classes of FPs which exhibit photo-transformable properties including photoactivation (PA), irreversible photo-conversion (PC) and reversible photo-switching (RS) [2]. Currently, irreversible photoconvertible and reversible photo-switchable FPs attract wide interest of scientists due to their potential of converting from an emissive state to another emissive state or switching between a fluorescent on and a non-fluorescent off state, respectively. The combination of reversibly switchable behavior and spectrally different emission has enabled application of multicolored super-resolution microscopy techniques in live-cell imaging. However, various drawbacks of currently used reversibly switchable red FPs (rsRFPs) have limited their application greatly and made them still being the least used in GFP-like proteins family. Moreover, the structure-function relationship and the mechanism controlling photo-switching behavior of rsRFPs have not been understood completely. Therefore, structural studies are essential to provide valuable information for the rational design of improved rsRFPs which fit better to experimental requirements.

The rsCherry protein was the first reported reversibly switchable red FP which was developed from mCherry – a good label in imaging techniques [3]. However, due to the non-optimal properties of rsCherry such as limited brightness, poor photostability and low contrast between on and off states [4]its application in super-resolution microscopy was not very widespread. Our current study has shown that rsCherry lost its absorption at 572 nm as well as fluorescence when it aged, despite being well protected from light, making studying its molecular structure and photo-mechanisms challenging. We were able to identify that the time-dependent bleaching in rsCherry is related to chromophore modifications and proposed that oxygen, a critical external reagent in the maturation process of FPs, is involved in unexpected chemical reactions of the chromophore. Spectroscopic data, native MS results and mutagenesis analysis, and especially structural studies of rsCherry crystallized in strictly anaerobic conditions strongly confirm our hypothesis that oxygen diminishes the rsCherry fluorescence through modifying its chromophore. These findings can help to develop improved red fluorescent proteins suitable for specific advanced imaging techniques.