In silico,in vitro,and in vivo machine learning in synthetic biology and metabolic engineering

Among the main learning methods reviewed in this study and used in synthetic biology and metabolic engineering are supervised learning, reinforcement and active learning, and in vitro or in vivo learning.In the context of biosynthesis, supervised machine learning is being exploited to predict biological sequence activities, predict structures and engineer sequences, and optimize culture conditions.Active and reinforcement learning methods use training sets acquired through an iterative process generally involving experimental measurements. They are applied to design, engineer, and optimize metabolic pathways and bioprocesses.The nascent but promising developments with in vitro and in vivo learning comprise molecular circuits performing simple tasks such as pattern recognition and classification.     

Quantitative analysis of the fermentative metabolism of glycerol in Escherichia coli

Availability, low price, and high degree of reduction have made glycerol a highly attractive and exploited carbon source for the production of fuels and reduced chemicals. Here we report the quantitative analysis of the fermentative metabolism of glycerol in Escherichia coli through the use of kinetic modeling and metabolic control analysis (MCA) to gain a better understanding of glycerol fermentation and identify key targets for genetic manipulation that could enhance product synthesis. The kinetics of glycerol fermentation in a batch culture was simulated using a dynamic model consisting of mass balances for glycerol, ethanol, biomass, and 11 intracellular metabolites, along with the corresponding kinetic expressions for the metabolism of each species. The model was then used to calculate metabolic control coefficients and elucidate the control structure of the pathways involved in glycerol utilization and ethanol synthesis. The calculated flux control coefficients indicate that the glycolytic flux during glycerol fermentation is almost exclusively controlled by the enzymes glycerol dehydrogenase (encoded by gldA) and dihydroxyacetone kinase (DHAK) (encoded by dhaKLM). In agreement with the MCA findings, overexpression of gldA and dhaKLM led to significant increase in glycerol utilization and ethanol synthesis fluxes. Moreover, overexpression of other enzymes involved in the pathways that mediate glycerol utilization and its conversion to ethanol had no significant impact on glycerol utilization and ethanol synthesis, further validating the MCA predictions. These findings were then applied as a means of increasing the production of ethanol: overexpression of glycerol dehyrdogenase and DHAK enabled the production of 20 g/L ethanol from crude glycerol, a by-product of biodiesel production, indicating the potential for industrial scale conversion of waste glycerol to ethanol under anaerobic conditions.     

Reverse genetics of Escherichia coli glycerol kinase allosteric regulation and glucose control of glycerol utilization in vivo

Reverse genetics is used to evaluate the roles in vivo of allosteric regulation of Escherichia coli glycerol kinase by the glucose-specific phosphocarrier of the phosphoenolpyruvate:glycose phosphotransferase system, IIA(Glc) (formerly known as III(glc)), and by fructose 1,6-bisphosphate. Roles have been postulated for these allosteric effectors in glucose control of both glycerol utilization and expression of the glpK gene. Genetics methods based on homologous recombination are used to place glpK alleles with known specific mutations into the chromosomal context of the glpK gene in three different genetic backgrounds. The alleles encode glycerol kinases with normal catalytic properties and specific alterations of allosteric regulatory properties, as determined by in vitro characterization of the purified enzymes. The E. coli strains with these alleles display the glycerol kinase regulatory phenotypes that are expected on the basis of the in vitro characterizations. Strains with different glpR alleles are used to assess the relationships between allosteric regulation of glycerol kinase and specific repression in glucose control of the expression of the glpK gene. Results of these studies show that glucose control of glycerol utilization and glycerol kinase expression is not affected by the loss of IIA(Glc) inhibition of glycerol kinase. In contrast, fructose 1,6-bisphosphate inhibition of glycerol kinase is the dominant allosteric control mechanism, and glucose is unable to control glycerol utilization in its absence. Specific repression is not required for glucose control of glycerol utilization, and the relative roles of various mechanisms for glucose control (catabolite repression, specific repression, and inducer exclusion) are different for glycerol utilization than for lactose utilization.     

In silico predictions of Escherichia coli metabolic capabilities are consistent with experimental data

A significant goal in the post-genome era is to relate the annotated genome sequence to the physiological functions of a cell. Working from the annotated genome sequence, as well as biochemical and physiological information, it is possible to reconstruct complete metabolic networks. Furthermore, computational methods have been developed to interpret and predict the optimal performance of a metabolic network under a range of growth conditions. We have tested the hypothesis that Escherichia coli uses its metabolism to grow at a maximal rate using the E. coli MG1655 metabolic reconstruction. Based on this hypothesis, we formulated experiments that describe the quantitative relationship between a primary carbon source (acetate or succinate) uptake rate, oxygen uptake rate, and maximal cellular growth rate. We found that the experimental data were consistent with the stated hypothesis, namely that the E. coli metabolic network is optimized to maximize growth under the experimental conditions considered. This study thus demonstrates how the combination of in silico and experimental biology can be used to obtain a quantitative genotype-phenotype relationship for metabolism in bacterial cells.     

A single amino acid change in Escherichia coli glycerol kinase abolishes glucose control of glycerol utilization in vivo.

Escherichia coli glycerol kinase (EC 2.7.1.30; ATP:glycerol 3-phosphotransferase) is a key element in glucose control of glycerol metabolism. Its catalytic activity is inhibited allosterically by the glycolytic intermediate, fructose 1,6-biphosphate, and by the phosphotransferase system phosphocarrier protein, IIIGlc (also known as IIAGlc). These inhibitors provide mechanisms by which glucose blocks glycerol utilization in vivo. We report here the cloning and sequencing of the glpK22 gene isolated from E. C. C. Lin strain 43, a strain that shows the loss of glucose control of glycerol utilization. DNA sequencing shows a single missense mutation that translates to the amino acid change Gly-304 to Ser (G-304-S) in glycerol kinase. The effects of this substitution on the functional and physical properties of the purified mutant enzyme were determined. Neither of the allosteric ligands inhibits it under conditions that produce strong inhibition of the wild-type enzyme, which is sufficient to explain the phenotype of strain 43. However, IIIGlc activates the mutant enzyme, which could not be predicted from the phenotype. In the wild-type enzyme, G-304 is located 1.3 nm from the active site and 2.5 nm from the IIIGlc binding site (M. Feese, D. W. Pettigrew, N. D. Meadow, S. Roseman, and S. J. Remington, Proc. Natl. Acad. Sci. USA 91:3544-3548, 1994). It is located in the same region as amino acid substitutions in the related protein DnaK which alter its catalytic and regulatory properties and which are postulated to interfere with a domain closure motion (A. S. Kamath-Loeb, C. Z. Lu, W.-C. Suh, M. A. Lonetto, and C. A. Gross, J. Biol. Chem. 270:30051-30059, 1995). The global effect of the G-304-S substitution on the conformation and catalytic and regulatory properties of glycerol kinase is consistent with a role for the domain closure motion in the molecular mechanism for glucose control of glycerol utilization.     

In silico deletion of PtsG gene in Escherichia coli genome-scale model predicts increased succinate production from glycerol

Systems metabolic engineering and in silico analyses are necessary to study gene knockout candidate for enhanced succinic acid production by Escherichia coli. Metabolically engineered E. coli has been reported to produce succinate from glucose and glycerol. However, investigation on in silico deletion of ptsG/b1101 gene in E. coli from glycerol using minimization of metabolic adjustment algorithm with the OptFlux software platform has not yet been elucidated. Herein we report what is to our knowledge the first direct predicted increase in succinate production following in silico deletion of the ptsG gene in E. coli GEM from glycerol with the OptFlux software platform. The result indicates that the deletion of this gene in E. coli GEM predicts increased succinate production that is 20% higher than the wild-type control model. Hence, the mutant model maintained a growth rate that is 77% of the wild-type parent model. It was established that knocking out of the ptsG/b1101 gene in E. coli using glucose as substrate enhanced succinate production, but the exact mechanism of this effect is still obscure. This study informs other studies that the deletion of ptsG/b1101 gene in E. coli GEM predicted increased succinate production, enabling a model-driven experimental inquiry and/or novel biological discovery on the underground metabolic role of this gene in E. coli central metabolism in relation to increasing succinate production when glycerol is the substrate.     

In silico reconstruction of the amino acid metabolic pathways of Trypanosoma cruzi

Trypanosoma cruzi is the epidemiological agent of Chagas' disease, affecting most of Central and South America, constituting a significant health and socio-economic problem. The parasite has a metabolism largely based on the consumption of amino acids, which participate in a diversity of metabolic pathways, leading to many crucial compounds for the survival of this parasite. Study of its enzymes has the potential to disclose new therapeutic targets and foster the development of new drugs. In this study, we employed computational approaches to reconstruct in silico the amino acid metabolic pathways of T. cruzi, aiming to link genomic information with functional information. For that, protein sequences from 570 EC classes belonging to 25 different pathways in general amino acid metabolism were downloaded from KEGG. A subset of 471 EC classes had at least one sequence deposited. Clustering of the proteins belonging to each EC class was performed using a similarity-based approach implemented in the tool AnEnPi. Reconstruction of the metabolic pathways comprising the amino acid metabolism of T. cruzi was performed by analyzing the output of BLASTP, using as query the dataset of predicted proteins of T. cruzi against all sequences of each individual cluster. This approach allowed us to identify 764 T. cruzi proteins probably involved in the metabolism of amino acids as well as the identification of several putative cases of analogy. Furthermore, we were able to identify several enzymatic activities of T. cruzi that were not previously included in KEGG.     

Redistribution of metabolic fluxes in Escherichia coli with fermentative lactate dehydrogenase overexpression and deletion

Under anaerobic conditions, competition for pyruvate between the branch point enzymes pyruvate formate lyase (PFL, Km = 2 mM) and fermentative lactate dehydrogenase (LDH, Km = 7.2 mM) determines the partition of carbon flux. Two Escherichia coli mutant strains, one deficient in ackA, pta, and ldhA and the other overexpressing LDH, were constructed to systematically analyze the effects of these perturbations in the existing pathways on the redistribution of carbon fluxes. Deletion of the lactate and acetate synthesis pathways was detrimental to cell growth. Carbon flux is forced through ethanol and formate production pathways, resulting in a concomitant increase in those fluxes. In addition, overexpression of LDH simultaneously increases the common flux as well as the flux to the competing acetyl-CoA branch. Overexpression of lactate dehydrogenase (ldhA) in the parent strain increases the lactate synthesis rate from 0.19 to 0.40 mmol/g-biomass-h when the LDH activities increases from 1.3 to 15.3 units. Even an increase of more than 10 times in the LDH activity fails to divert a large fraction of the carbon flux to lactate; the majority of the flux still channels through the acetyl-CoA branch. Overexpression of LDH in the parent strain simultaneously increases the common flux as well as the flux through the acetyl-CoA branch. Subsequently, the flux amplification factors (or deviation indices which can be related to the flux control coefficients) are positive for all three fluxes occurring at the pyruvate node.     

In silico design of anaerobic growth-coupled product formation in Escherichia coli: experimental validation using a simple polyol,glycerol

Bioprocess and Biosystems Engineering - Integrated approaches using in silico model-based design and advanced genetic tools have enabled efficient production of fuels, chemicals and functional...     

In vivo effects of pyrophosphate in Escherichia coli

Abstract Pyrophosphate (PP_i), a noncompetitive inhibitor of Rho poly(C)-dependent ATPase activity in vitro has been shown to relieve polarity in the lac operon. This indicates that PP_i could inhibit Rho activity in vivo too. An additional effect of PP_i on adenosine 3',5'-cyclic monophosphate (cAMP) synthesis during stationary phase of growth is also described.     

In vivo,in vitro and in silico: an open space for the development of microbe-based applications of synthetic biology

Living systems are studied using three complementary approaches: living cells, cell-free systems and computer-mediated modelling. Progresses in understanding, allowing researchers to create novel chassis and industrial processes rest on a cycle that combines in vivo, in vitro and in silico studies. This design–build–test–learn iteration loop cycle between experiments and analyses combines together physiology, genetics, biochemistry and bioinformatics in a way that keeps going forward. Because computer-aided approaches are not directly constrained by the material nature of the entities of interest, we illustrate here how this virtuous cycle allows researchers to explore chemistry which is foreign to that present in extant life, from whole chassis to novel metabolic cycles. Particular emphasis is placed on the importance of evolution.     

The evolution and structural anatomy of the small molecule metabolic pathways in Escherichia coli

The 106 small molecule metabolic (SMM) pathways in Escherichia coli are formed by the protein products of 581 genes. We can define 722 domains, nearly all of which are homologous to proteins of known structure, that form all or part of 510 of these proteins. This information allows us to answer general questions on the structural anatomy of the SMM pathway proteins and to trace family relationships and recruitment events within and across pathways. Half the gene products contain a single domain and half are formed by combinations of between two and six domains. The 722 domains belong to one of 213 families that have between one and 51 members. Family members usually conserve their catalytic or cofactor binding properties; substrate recognition is rarely conserved. Of the 213 families, members of only a quarter occur in isolation, i.e. they form single-domain proteins. Most members of the other families combine with domains from just one or two other families and a few more versatile families can combine with several different partners.Excluding isoenzymes, more than twice as many homologues are distributed across pathways as within pathways. However, serial recruitment, with two consecutive enzymes both being recruited to another pathway, is rare and recruitment of three consecutive enzymes is not observed. Only eight of the 106 pathways have a high number of homologues. Homology between consecutive pairs of enzymes with conservation of the main substrate-binding site but change in catalytic mechanism (which would support a simple model of retrograde pathway evolution) occurs only six times in the whole set of enzymes. Most of the domains that form SMM pathways have homologues in non-SMM pathways. Taken together, these results imply a pervasive "mosaic" model for the formation of protein repertoires and pathways.     

In vivo dissection of the Tat translocation pathway in Escherichia coli

The bacterial Tat pathway is capable of exporting folded proteins carrying a special twin arginine (RR) signal peptide. By using two in vivo reporter proteins, we assessed factors that affect Tat pathway transport. We observed that, like the intact RR signal peptide, those with a KR or RK substitution were still capable of mediating the translocation of the folded green fluorescent protein (GFP). However, the translocation efficiency decreased in the order of RR>KR>RK. The KK motif was unable to mediate GFP translocation. The translocation of the RR-GFP fusion required TatA, TatB and TatC proteins. By exploiting the periplasmic bactericidal property of colicin V (ColV), we constructed a translocation-suicide probe, RR-ColV. The translocation of RR-ColV fully inhibited the growth of wild-type Escherichia coli and those of the DeltatatD and DeltatatE mutants. In contrast, the deletion of the tatC gene blocked RR-ColV in the cytoplasm and this strain exhibited a normal growth phenotype. Interestingly, the growth of DeltatatA and tatB mutants was inhibited partially by RR-ColV. Moreover, KR, RK and KK motifs were capable of mediating the ColV translocation with a decreasing RR=KR>RK>KK efficiency. In addition to TatE and TatC proteins, either TatA or TatB was sufficient for the translocation of RR-ColV or KR-ColV. In contrast, TatA plus the conserved N-terminal domain of TatB were required to mediate the killing effect of ColV fused to the less-efficient RK signal peptide. Taken together, these results suggest that a fully efficient Tat pathway transport is determined by the sequence of the signal peptide, the composition of the Tat apparatus, and the intrinsic characteristics of exported proteins.