Evolution-based design of chorismate mutase enzymes

Russ, William P.; Figliuzzi, Matteo; Stöcker, Christian; Barrat-Charlaix, Pierre; Socolich, Michael; Kast, Peter; Hilvert, Donald; Monasson, Rémi; Cocco, Simona; Weigt, Martin; Ranganathan, Rama

doi:10.1101/2020.04.01.020487

Cited by 26 publications

(45 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…7). A similar upper value for kcat/Km was recently confirmed for a set of previously unknown CMs that were sampled from 1130 natural AroQ sequences of phylogenetically widely diverse organisms (46). The systematic decrease of improvements per evolutionary round and the inability to go beyond ≈10 6 M -1 s -1 in spite of applying sophisticated evolutionary strategies might indicate that we have approached an intrinsic threshold for the evolution of the enzymatic Claisen rearrangement of chorismate.…”

Section: Discussionsupporting

confidence: 75%

Evolving the naturally compromised chorismate mutase from Mycobacterium tuberculosis to top performance

Fahrig-Kamarauskait≑

Würth-Roderer

Thorbjornsrud

et al. 2020

Journal of Biological Chemistry

Self Cite

View full text Add to dashboard Cite

Chorismate mutase (CM), an essential enzyme at the branch point of the shikimate pathway, is required for the biosynthesis of phenylalanine and tyrosine in bacteria, archaea, plants, and fungi. MtCM, the CM from Mycobacterium tuberculosis, has less than 1% of the catalytic efficiency of a typical natural CM and requires complex formation with DAHP synthase for high activity. To explore the full potential of MtCM for catalyzing its native reaction, we applied diverse iterative cycles of mutagenesis and selection, thereby raising kcat/Km 270-fold to 5 × 105 M-1s-1, which is even higher than for the complex. Moreover, the evolutionarily optimized autonomous MtCM, which had 11 of its 90 amino acids exchanged, was stabilized compared to its progenitor, as indicated by a 9 °C increase in melting temperature. The 1.5 Å crystal structure of the top-evolved MtCM variant reveals the molecular underpinnings of this activity boost. Some acquired residues, (e.g. Pro52 and Asp55) are conserved in naturally efficient CMs, but most of them lie beyond the active site. Our evolutionary trajectories reached a plateau at the level of the best natural enzymes suggesting that we have exhausted the potential of MtCM. Taken together, these findings show that the scaffold of MtCM, which naturally evolved for mediocrity to enable inter-enzyme allosteric regulation of the shikimate pathway, is inherently capable of high activity.

show abstract

Section: Discussionsupporting

confidence: 75%

Evolving the naturally compromised chorismate mutase from Mycobacterium tuberculosis to top performance

Fahrig-Kamarauskait≑

Würth-Roderer

Thorbjornsrud

et al. 2020

Journal of Biological Chemistry

Self Cite

View full text Add to dashboard Cite

show abstract

“…Additionally, our two part mechanism to explain this performance provides context for and extends previous unsupervised protein function modeling and design work. While unsupervised methods trained on natural sequence data perform well at predicting or avoiding loss-of-function variants during modeling and design, they have also been unable to reliably model or design better-than-natural variants 38,39,41,47,52 . Our findings suggest that a small amount of labeled data and additional supervised learning on top of unsupervised pre-training may be necessary to find enhanced variants.…”

Section: Discussionmentioning

confidence: 99%

“…As these analyses only consider simple "first-order" mutational overlap, they provide a lower-bound on non-triviality. Indeed, due to epistasis, even recombining existing mutations among evolutionary homologs to produce functional proteins is a difficult challenge 47 .…”

Section: Low-n Engineering Of the Fluorescent Protein Avgfpmentioning

confidence: 99%

Low-N protein engineering with data-efficient deep learning

Biswas

Khimulya

Alley

et al. 2020

Preprint

190

View full text Add to dashboard Cite

Protein engineering has enormous academic and industrial potential. However, it is limited by the lack of experimental assays that are consistent with the design goal and sufficiently high-throughput to find rare, enhanced variants. Here we introduce a machine learning-guided paradigm that can use as few as 24 functionally assayed mutant sequences to build an accurate virtual fitness landscape and screen ten million sequences via in silico directed evolution. As demonstrated in two highly dissimilar proteins, avGFP and TEM-1 β-lactamase, top candidates from a single round are diverse and as active as engineered mutants obtained from previous multi-year, high-throughput efforts. Because it distills information from both global and local sequence landscapes, our model approximates protein function even before receiving experimental data, and generalizes from only single mutations to propose high-functioning epistatically non-trivial designs. With reproducible >500% improvements in activity from a single assay in a 96-well plate, we demonstrate the strongest generalization observed in machine-learning guided protein design to date. Taken together, our approach enables efficient use of resource intensive high-fidelity assays without sacrificing throughput. By encouraging alignment with endpoint objectives, low-N design will accelerate engineered proteins into the fermenter, field, and clinic.

show abstract

“…2C-E). These generated peptides could be experimentally tested in future, providing additional evidence for the validity of the model [32].…”

Section: Discussionmentioning

confidence: 99%

Flexible machine learning prediction of antigen presentation for rare and common HLA-I alleles

Bravi

Tubiana

Cocco

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

The recent increase of immunopeptidomic data, obtained by mass spectrometry (MS) or binding assays, opens unprecedented possibilities for investigating endogenous antigen presentation by the highly polymorphic human leukocyte antigen class I (HLA-I) protein. We introduce a flexible and easily interpretable peptide presentation prediction method. We validate its performance as a predictor of cancer neoantigens and viral epitopes and use it to reconstruct peptide motifs presented on specific HLA-I molecules.Recognition of malignant and infected cells by the adaptive immune system requires binding of cytotoxic T-cell receptors to antigens, 8-11-mer peptides presented by the Major Histocompatibilty Complex (MHC) class I coded by HLA-I alleles ( Fig. 1a). Tumour-specific neoantigens are currently sought-after targets for improving cancer immunotherapy [1, 2]. Computational predictions can help select potential neoantigens and accelerate immunogenicity testing. To be useful, these predictions must be specific to each HLA type.State-of-the-art methods [3][4][5], such as NetMHC [6, 7], are based on artificial neural networks trained in a supervised way to predict peptide presentation from known peptide-HLA association. They must be trained on large datasets, and perform best on common alleles. Their accuracy is degraded for rare or little studied HLA-I alleles which are poorly represented in databases. In that case, another approach is to train unsupervised models of presentation from custom elution experiments with little or no information about peptide-HLA association. For instance, MixMHCp [8, 9] can reconstruct, from unannotated peptide sequences, a mixture of generative models -one for each expressed HLA type. However, it makes simplifying assumptions about binding specificity, and is not designed to leverage available (albeit limited) annotation information from the Immune Epitope Database [10] (IEDB) to improve accuracy.We present an alternative method for predicting peptides presented by specific HLA types, which can be trained on custom datasets. It can be applied to patient-or experiment-specific samples and treats common and rare alleles on equal footing, avoiding potential database biases. We use a Restricted Boltzmann Machine (RBM), an unsupervised machine learning scheme that learns probability distributions of sequences given as input [11][12][13]. The RBM estimates presentation scores for each peptide, and can generate candidate presentable peptides. It also provides a lower dimensional representation of peptides with a clear interpretation in terms of associated HLA type, which can be exploited to train classifiers of HLA association from a small number of HLA annotations. The RBM has a simple structure with one hidden layer and weights connecting the input -the peptide sequence -to the hidden layer. The weights, along with the biases acting on both input and hidden units, are learned from a list of presented peptides (see Methods, Fig. 1b, Supplementary Figs. 1, 2, 3). The RBM probability is interprete...

show abstract

Evolution-based design of chorismate mutase enzymes

Cited by 26 publications

References 40 publications

Evolving the naturally compromised chorismate mutase from Mycobacterium tuberculosis to top performance

Evolving the naturally compromised chorismate mutase from Mycobacterium tuberculosis to top performance

Low-N protein engineering with data-efficient deep learning

Flexible machine learning prediction of antigen presentation for rare and common HLA-I alleles

Contact Info

Product

Resources

About