Machine Learning Enables Selection of Epistatic Enzyme Mutants for Stability Against Unfolding and Detrimental Aggregation.

Machine learning (ML) has pervaded most areas of protein engineering, including stability and stereoselectivity. Using limonene epoxide hydrolase as the model enzyme and innov'SAR as the ML platform, comprising a digital signal process, we achieved high protein robustness that can resist unfolding with concomitant detrimental aggregation. Fourier transform (FT) allows us to take into account the order of the protein sequence and the nonlinear interactions between positions, and thus to grasp epistatic phenomena. The innov'SAR approach is interpolative, extrapolative and makes outside-the-box, predictions not found in other state-of-the-art ML or deep learning approaches. Equally significant is the finding that our approach to ML in the present context, flanked by advanced molecular dynamics simulations, uncovers the connection between epistatic mutational interactions and protein robustness.

One-shot optimization of multiple enzyme parameters: Tailoring glucose oxidase for pH and electron mediators.

Enzymes are biological catalysts with many industrial applications, but natural enzymes are usually unsuitable for industrial processes because they are not optimized for the process conditions. The properties of enzymes can be improved by directed evolution, which involves multiple rounds of mutagenesis and screening. By using mathematical models to predict the structure-activity relationship of an enzyme, and by defining the optimal combination of mutations in silico, we can significantly reduce the number of bench experiments needed, and hence the time and investment required to develop an optimized product. Here, we applied our innovative sequence-activity relationship methodology (innov'SAR) to improve glucose oxidase activity in the presence of different mediators across a range of pH values. Using this machine learning approach, a predictive model was developed and the optimal combination of mutations was determined, leading to a glucose oxidase mutant (P1) with greater specificity for the mediators ferrocene-methanol (12-fold) and nitrosoaniline (8-fold), compared to the wild-type enzyme, and better performance in three pH-adjusted buffers. The kcat /KM ratio of P1 increased by up to 121 folds compared to the wild type enzyme at pH 5.5 in the presence of ferrocene methanol.

A machine learning approach for reliable prediction of amino acid interactions and its application in the directed evolution of enantioselective enzymes.

Directed evolution is an important research activity in synthetic biology and biotechnology. Numerous reports describe the application of tedious mutation/screening cycles for the improvement of proteins. Recently, knowledge-based approaches have facilitated the prediction of protein properties and the identification of improved mutants. However, epistatic phenomena constitute an obstacle which can impair the predictions in protein engineering. We present an innovative sequence-activity relationship (innov'SAR) methodology based on digital signal processing combining wet-lab experimentation and computational protein design. In our machine learning approach, a predictive model is developed to find the resulting property of the protein when the n single point mutations are permuted (2n combinations). The originality of our approach is that only sequence information and the fitness of mutants measured in the wet-lab are needed to build models. We illustrate the application of the approach in the case of improving the enantioselectivity of an epoxide hydrolase from Aspergillus niger. n = 9 single point mutants of the enzyme were experimentally assessed for their enantioselectivity and used as a learning dataset to build a model. Based on combinations of the 9 single point mutations (29), the enantioselectivity of these 512 variants were predicted, and candidates were experimentally checked: better mutants with higher enantioselectivity were indeed found.

In vivo evolution of metabolic pathways: Assembling old parts to build novel and functional structures.

In our recent article "In vivo evolution of metabolic pathways by homeologous recombination in mitotic cells" we proposed a useful alternative to directed evolution methods that permits the generation of yeast cell libraries containing recombinant metabolic pathways from counterpart genes. The methodology was applied to generate single mosaic genes and intragenic mosaic pathways. We used flavonoid metabolism genes as a working model to assembly and express evolved pathways in DNA repair deficient cells. The present commentary revises the principles of gene and pathway mosaicism and explores the scope and perspectives of our results as an additional tool for synthetic biology.

In vivo evolution of metabolic pathways by homeologous recombination in mitotic cells.

We describe a rapid and highly efficient method for the assembly, recombination, targeted chromosomal integration and regulatable expression of mosaic metabolic pathways by homeologous recombination in DNA repair deficient yeast cells. We have assembled and recombined 23kb pathways containing all the genes encoding enzymes for the production of flavonoids, a group of plant secondary metabolites of nutritional and agricultural value. The mosaic genes of the pathways resulted from pair-wise recombination of two nonidentical (homeologous) wild-type genes. The recombination events occurred simultaneously in the cell. Correctly assembled mosaic gene clusters could only be observed in DNA repair deficient strains. Thus, libraries of intragenic mosaic pathways were generated. Randomly isolated clones were screened for their ability to produce flavonoids such as kaempferol, phloretin and galangin. Thus, the functionality of the recombinant pathways was proven. Additionally, significant higher concentrations of metabolites such as naringenin, pinocembrin and dihydrokaempferol were detected. Further analysis also revealed the production of different aromatic compounds such as styrene, hydroxystyrene, phloretic acid and other molecules. We show that the in vivo homeologous recombination strategy can generates libraries of intragenic mosaic pathways producing a high diversity of phenylpropanoid compounds.

MICAL-like1 mediates epidermal growth factor receptor endocytosis.

Small GTPase Rabs are required for membrane protein sorting/delivery to precise membrane domains. Rab13 regulates epithelial tight junction assembly and polarized membrane transport. Here we report that Molecule Interacting with CasL (MICAL)-like1 (MICAL-L1) interacts with GTP-Rab13 and shares a similar domain organization with MICAL. MICAL-L1 has a calponin homology (CH), LIM, proline rich and coiled-coil domains. It is associated with late endosomes. Time-lapse video microscopy shows that green fluorescent protein-Rab7 and mcherry-MICAL-L1 are present within vesicles that move rapidly in the cytoplasm. Depletion of MICAL-L1 by short hairpin RNA does not alter the distribution of a late endosome/lysosome-associated protein but affects the trafficking of epidermal growth factor receptor (EGFR). Overexpression of MICAL-L1 leads to the accumulation of EGFR in the late endosomal compartment. In contrast, knocking down MICAL-L1 results in the distribution of internalized EGFR in vesicles spread throughout the cytoplasm and promotes its degradation. Our data suggest that the N-terminal CH domain associates with the C-terminal Rab13 binding domain (RBD) of MICAL-L1. The binding of Rab13 to RBD disrupts the CH/RBD interaction, and may induce a conformational change in MICAL-L1, promoting its activation. Our results provide novel insights into the MICAL-L1/Rab protein complex that can regulate EGFR trafficking at late endocytic pathways.

The yeast zinc finger regulators Pdr1p and Pdr3p control pleiotropic drug resistance (PDR) as homo- and heterodimers in vivo.

The transcription factors Pdr1p and Pdr3p from Saccharomyces cerevisiae mediate pleiotropic drug resistance (PDR) by controlling expression of ATP-binding cassette (ABC) transporters such as Pdr5p, Snq2p and Yor1p. Previous in vitro studies demonstrated that Pdr1p and Pdr3p recognize so-called pleiotropic drug resistance elements (PDREs) in the promoters of target genes. In this study, we show that both Pdr1p and Pdr3p are phosphoproteins; Pdr3p isoforms migrate as two bands in gel electrophoresis, reflecting two distinct phosphorylation states. Most importantly, native co-immunoprecipitation experiments, using functional epitope-tagged Pdr1p/Pdr3p variants, demonstrate that Pdr1p and Pdr3p can form both homo- and heterodimers in vivo. Furthermore, in vivo footprinting of PDRE-containing promoters demonstrate that Pdr1p/Pdr3p constitutively occupy both perfect and degenerate PDREs in vivo. Thus, in addition to interaction with other regulators, differential dimerization provides a plausible explanation for the observation that Pdr3p and Pdr1p can both positively and negatively control PDR promoters with different combinations of perfect and degenerate PDREs.

The small GTPase Rab13 regulates assembly of functional tight junctions in epithelial cells.

Junctional complexes such as tight junctions (TJ) and adherens junctions are required for maintaining cell surface asymmetry and polarized transport in epithelial cells. We have shown that Rab13 is recruited to junctional complexes from a cytosolic pool after cell-cell contact formation. In this study, we investigate the role of Rab13 in modulating TJ structure and functions in epithelial MDCK cells. We generate stable MDCK cell lines expressing inactive (T22N mutant) and constitutively active (Q67L mutant) Rab13 as GFP-Rab13 chimeras. Expression of GFP-Rab13Q67L delayed the formation of electrically tight epithelial monolayers as monitored by transepithelial electrical resistance (TER) and induced the leakage of small nonionic tracers from the apical domain. It also disrupted the TJ fence diffusion barrier. Freeze-fracture EM analysis revealed that tight junctional structures did not form a continuous belt but rather a discontinuous series of stranded clusters. Immunofluorescence studies showed that the expression of Rab13Q67L delayed the localization of the TJ transmembrane protein, claudin1, at the cell surface. In contrast, the inactive Rab13T22N mutant did not disrupt TJ functions, TJ strand architecture nor claudin1 localization. Our data revealed that Rab13 plays an important role in regulating both the structure and function of tight junctions.