Implications of AlphaFold2 for crystallographic phasing by molecular replacement

The AlphaFold2 results in the 14th edition of Critical Assessment of Structure Prediction (CASP14) showed that accurate (low root-mean-square deviation) in silico models of protein structure domains are on the horizon, whether or not the protein is related to known structures through high-coverage sequence similarity. As highly accurate models become available, generated by harnessing the power of correlated mutations and deep learning, one of the aspects of structural biology to be impacted will be methods of phasing in crystallography. Here, the data from CASP14 are used to explore the prospects for changes in phasing methods, and in particular to explore the prospects for molecular-replacement phasing using in silico models.

De novo determination of mosquitocidal Cry11Aa and Cry11Ba structures from naturally-occurring nanocrystals

Cry11Aa and Cry11Ba are the two most potent toxins produced by mosquitocidal Bacillus thuringiensis subsp. israelensis and jegathesan, respectively. The toxins naturally crystallize within the host; however, the crystals are too small for structure determination at synchrotron sources. Therefore, we applied serial femtosecond crystallography at X-ray free electron lasers to in vivo-grown nanocrystals of these toxins. The structure of Cry11Aa was determined de novo using the single-wavelength anomalous dispersion method, which in turn enabled the determination of the Cry11Ba structure by molecular replacement. The two structures reveal a new pattern for in vivo crystallization of Cry toxins, whereby each of their three domains packs with a symmetrically identical domain, and a cleavable crystal packing motif is located within the protoxin rather than at the termini. The diversity of in vivo crystallization patterns suggests explanations for their varied levels of toxicity and rational approaches to improve these toxins for mosquito control.

Artificial intelligence to solve the X-ray crystallography phase problem: a case study report.

The determination of three dimensional structures of macromolecules is one of the actual challenge in biology with the ultimate objective of understanding their function. So far, X-ray crystallography is the most popular method to solve structure, but this technique relies on the generation of diffracting crystals. Once a correct data set has been obtained, the calculation of electron density maps requires to solve the so-called phase problem using different approaches. The most frequently used technique is molecular replacement, which relies on the availability of the structure of a protein sharing strong structural similarity with the studied protein. Its success rate is directly correlated with the quality of the models used for the molecular replacement trials. The availability of models as accurate as possible is then definitely critical. Very recently, a breakthrough step has been made in the field of protein structure prediction thanks to the use of machine learning approaches as implemented in the AlphaFold or RoseTTAFold structure prediction programs. Here, we describe how these recent improvements helped us to solve the crystal structure of a protein involved in the nonsense-mediated mRNA decay pathway (NMD), an mRNA quality control pathway dedicated to the elimination of eukaryotic mRNAs harboring premature stop codons.

Improved resolution crystal structure of Acanthamoeba actophorin reveals structural plasticity not induced by microgravity

Actophorin, a protein that severs actin filaments isolated from the amoeba Acanthamoeba castellanii, was employed as a test case for crystallization under microgravity. Crystals of purified actophorin were grown under microgravity conditions aboard the International Space Station (ISS) utilizing an interactive crystallization setup between the ISS crew and ground-based experimenters. Crystals grew in conditions similar to those grown on earth. The structure was solved by molecular replacement at a resolution of 1.65 Å. Surprisingly, the structure reveals conformational changes in a remote β-turn region that were previously associated with actophorin phosphorylated at the terminal residue Ser1. Although crystallization under microgravity did not yield a higher resolution than crystals grown under typical laboratory conditions, the conformation of actophorin obtained from solving the structure suggests greater flexibility in the actophorin β-turn than previously appreciated and may be beneficial for the binding of actophorin to actin filaments.

AlphaFold Protein Structure Database for Sequence-Independent Molecular Replacement

Crystallographic phasing recovers the phase information that is lost during a diffraction experiment. Molecular replacement is a commonly used phasing method for crystal structures in the protein data bank. In one form it uses a protein sequence to search a structure database to find suitable templates for phasing. However, sequence information is not always available, such as when proteins are crystallized with unknown binding partner proteins or when the crystal is of a contaminant. The recent development of AlphaFold published the predicted protein structures for every protein from twenty distinct species. In this work, we tested whether AlphaFold-predicted E. coli protein structures were accurate enough to enable sequence-independent phasing of diffraction data from two crystallization contaminants of unknown sequence. Using each of more than 4000 predicted structures as a search model, robust molecular replacement solutions were obtained, which allowed the identification and structure determination of YncE and YadF. Our results demonstrate the general utility of the AlphaFold-predicted structure database with respect to sequence-independent crystallographic phasing.

AlphaFold Protein Structure Database for Sequence-Independent Molecular Replacement

Crystallographic phasing recovers the phase information that is lost during a diffraction experiment. Molecular replacement is a dominant phasing method for the crystal structures in the protein data bank. In one form it uses a protein sequence to search a structure database for finding suitable templates for phasing. However, such sequence information is not always available such as when proteins are crystallized with unknown binding partner proteins or when the crystal is that of a contaminant. The recent development of AlphaFold has resulted in the availability of predicted protein structures for all proteins from twenty species. In this work, we tested whether AlphaFold-predicted E. coli protein structures were accurate enough for sequence-independent phasing of diffraction data from two crystallization contaminants for which we had not identified the protein. Using each of more than 4000 predicted structures as a search model, robust molecular replacement solutions were obtained which allowed the identification and structure determination of both structures, YncE and YadF. Our results advocate a general utility of AlphaFold-predicted structure database with respect to crystallographic phasing.

MrParse: Finding homologues in the PDB and the EBI AlphaFold database for Molecular Replacement and more

Crystallographers have an array of search model options for structure solution by Molecular Replacement (MR). Well-established options of homologous experimental structures and regular secondary structure elements or motifs are increasingly supplemented by computational modelling. Such modelling may be carried out locally or use pre-calculated predictions retrieved from databases such as the EBI AlphaFold database. MrParse is a new pipeline to help streamline the decision process in MR by consolidating bioinformatic predictions in one place. When reflection data are provided, MrParse can rank any homologues found using eLLG which indicates the likelihood that a given search model will work in MR. In-built displays of predicted secondary structure, coiled-coil and transmembrane regions further inform the choice of MR protocol. MrParse can also identify and rank homologues in the EBI AlphaFold database, a function that will also interest other structural biologists and bioinformaticians.

Computational models in the service of X-ray and cryo-EM structure determination

CASP (Critical Assessment of Structure prediction) conducts community experiments to determine the state of the art in computing protein structure from amino acid sequence. The process relies on the experimental community providing information about not yet public or about to be solved structures, for use as targets. For some targets, the experimental structure is not solved in time for use in CASP. Calculated structure accuracy improved dramatically in this round, implying that models should now be much more useful for resolving many sorts of experimental difficulty. To test this, selected models for seven unsolved targets were provided to the experimental groups. These models were from the AlphaFold2 group, who overall submitted the most accurate predictions in CASP14. Four targets were solved with the aid of the models, and, additionally, the structure of an already solved target was improved. An a-posteriori analysis showed that in some cases models from other groups would also be effective. This paper provides accounts of the successful application of models to structure determination, including molecular replacement for X-ray crystallography, backbone tracing and sequence positioning in a Cryo-EM structure, and correction of local features. The results suggest that in future there will be greatly increased synergy between computational and experimental approaches to structure determination.