Team:Amsterdam/Results

Forbidden FRUITS

Overview

Industry is a significant source of environmental pollution - be it from the non-renewable substrates it typically relies on or from the waste streams it produces. Ideally, we would want the industrial processes on which our society depends to be as circular as possible, i.e. substrates that become products, become substrates again. Over the last couple of years, we have been looking for sustainable alternatives. Evermore pertinent, cell factories can serve to replace conventional production methods or use pollutants as an energy source to generate compounds or products for us. In order to accelerate the use of cell factories in industry to produce compounds more sustainably, we have developed an algorithm called Forbidden FRUITS, that can generate genetic strategies for production of compounds within a microorganism. Using reactions from multiple databases (Merger) and their association with specific genes and proteins (Network Transformation Method based on GPRA), we are able to predict producing pathways (Cheap lunch), couple the production pathway to growth (Plumber) and optimize the growth-coupled production pathway (Optimization using Downstream Metabolite Information). Experimentally, we will validate the growth-coupled genetic strategy for i) the production of salicylic acid in both Synechocystis PCC6803 and E. coli K-12; ii) for the production of lactate in Synechocystis PCC6803 and Synechococcus UTEX 2973; and iii) for the production of mannitol in Synechocystis PCC6803.

Key Takeaways

We have written a modular algorithm to growth-couple any product to any microorganism. We created the following modules which individually add unique features to Forbidden FRUITS:

Merger: exploits the complementary informations available in multiple online databases
Gene-Protein-Reaction Association network transformer: adds the relations between reactions and genes
Cheap lunch strategy finder: expands the variety of products which can be made by enabling indirect growth coupling
Plumber: finds modification to make product formation essential for biomass formation
Method for the optimization of strategies using downstream metabolite information: is able to increase product yield

We showed that Forbidden FRUITS is able to develop a chassis of different types of microorganisms for the production of a variety of compounds by implementing strategies for the following coupling metabolites:

Pyruvate (salicylic acid): in Synechocystis PCC6803 and in E.coli K-12
Oxaloacetate (lactate): in Synechocystis PCC6803 and Synechococcus UTEX 2793

We showed that we can use strategies computed using the cheap lunch method to make a larger range of products by implementing a strategy for the production of mannitol in Synechocystis PCC6803

Future prospects

A project never really finishes, there are always parts that could be improved. Forbidden FRUITS is not an exception. Here we list the future perspectives for Forbidden FRUITS

Incorporating regulatory networks

The algorithm could be improved by incorporating regulatory networks of example protein - protein interactions and gene regulatory networks. These networks can be used to optimize the strategy to knock-out and knock-in genes in the metabolic network.

Incorporating information about product transporters and diffusion rates

To check if the desired product can be easily obtained from the cell.

Incorporation of sequence optimizer

The non-native product formation reaction originates from other species than the microbe of choice. The gene sequence associated to this reaction is therefore not optimal for the host microorganism. A sequence optimizer could potentially adjust the sequence to a more native variant to the host bacteria.

Enabling choice of substrate

In order to use other organisms as sustainable cell factories, we could rewrite our methods such that we are also able to choose any substrate, thereby creating cell factories which can use waste products for growth and production.

Adjust Forbidden FRUITS modules to make them useful for other industries

Since we created the modules of Forbidden FRUITS such that they can be used in a flexible way, it is interesting to explore possible applications of some modules in different fields. The Plumber could for example be used in economics to define a minimal cost strategy for a production of an industrial product. The merger could be used to merge different databases for different types of data. The merger in combination with the plumber could also be used in chemical industries for the production of enzymes.

Also for the wetlab there are still some future perspectives:

Large-scale experiments

In order to show that stable production can be achieved by implementing strategies generated by Forbidden FRUITS, we could grow producing strains for a longer period of time in a more industrial setting.

Apply more Forbidden FRUITS strategies to different microorganisms and to produce more different compounds
Experimental validation of genetic strategies for more product-microorganism combinations that were generated by Forbidden FRUITS

We could try to implement to more strategies to produce secondary metabolites
When we obtain a stable pyruvate or lactate producing Synechocystis PCC6803 or Synechococcus UTEX 2973 strain, we could try to incorporate different production pathways to show that Forbidden FRUITS can be used to create a chassis for the production of several non-native compounds.

Dry lab

Database merger

By combining information from multiple sources, Forbidden FRUITS exploits the complementary information which is available in multiple online databases. This will extend the availability of compounds and reactions that the algorithm can use to find usable pathways and this will lead to the prediction of better strategies.

Overview

The predictions of Forbidden FRUITS are limited by the amount of information available to it. Currently biochemical information is scattered over many different databases, each with their own strengths and weaknesses. To combine information from multiple databases, we developed the merger. We designed a database format that would fit multiple databases and would allow for the putative incorporation of new databases. Since each database comes with its own format to store reactions, metabolites, genes and additional information, a parser was created to standardise the information obtained from each database. The goal of the merger is to create a final database containing a single entry for every reaction and compound while retaining as much meta information from the databases for possible later use. The central problem our merger faces is how to decide which reactions or compounds to merge. Since the amount of (meta-)information contained in each database is different, several methods were developed. For instance, the merger can use the reaction stoichiometry, the to find similar reactions and metabolites, which increases the ability of the merger to detect duplicate entries originating from different databases.

The merger was tested and reiterated using custom designed toy databases, which contain many of the possible exceptions that could be present in real databases. The finalized merger is able to load and merge JSON datafiles of the KEGG, BiGG, ModelSeed and MetaNetX databases, as well as SBML datafiles containing computational models of biological systems. In Forbidden FRUITS these are the metabolic networks of the microbe of choice.

For more details on the merger, visit our Engineering success page

Key takeaways:

We created a parsing method for the KEGG, BiGG, ModelSeed and MetaNetX databases to mold a raw database file into a flexible format
We created a merging method to combine multiple databases in a single file
We created a stoichiometric merging method, which uses reaction stoichiometry to merge databases.
Merging multiple databases results in a bigger and more complete dataset. Forbidden FRUITS uses this extra information to predict better strategies.

Network Transformation Method based on Gene-Protein-Reaction Associations

By adding gene-protein-reaction associations we are able to link the reactions in the network directly to gene knockouts. Incorporating gene-protein information allows Forbidden FRUITS to use information about isozymes and promiscuous genes, which will improve the feasibility of the generated designs.

Overview

We built a Gene-Protein-Reaction Associations (GPRA) parsing method to extract GPRA information fromSBML models. In order to use this information further downstream in the pipeline we developed a network transformation method, which is able to extend a metabolic network of reactions to a gene-based network while taking reaction reversibility into account.

For more details on the GPRA, visit our Engineering success page

We validated our method by performing parsimonious Flux Balance Analysis (pFBA) on the E. coli model iAF1260. For this model others have already modelled the fluxes through central carbon reactions using pFBA and GPRA information (Machado et.al, 2015). Our results (shown in Table 1), show that our method predicts the same fluxes and therefore works as expected.

**Table 1. pFBA results on iAF1260 with network transformation method.** F6PA: Fructose 6-phosphate aldolase, PFK: phosphofructokinase, FBA: Fructose-bisphosphate aldolase, PYK: Pyruvate kinase, DHAPT: Dihydroxyacetone phosphate transferase.
	On original network		On transformed network
Reaction/enzyme	Expected value	Obtained value	Expected value	Obtained value
F6PA	1.4	1.436	0.0	0.000
PFK	6.2	6.208	7.6	7.644
FBA	6.2	6.208	7.6	7.644
PYK	0.0	0.000	1.4	1.436
DHAPT	1.4	1.436	0.0	0.000

Key takeaways:

We built a GPRA parsing method, which can read GPRA information from SBML models.
We created a network transformation method, which transforms the stoichiometric matrix of reactions into irreversible reactions and extends each reaction with corresponding GPRA.
The network building and transformation method was validated using parsimonious Flux Balance Analysis (pFBA)

Cheap-lunch strategy-finder

Cheap-lunch strategy finder enables Forbidden FRUITS to find paths which couple the production of a specific compound indirectly to biomass formation. With the incorporation of these strategies to Forbidden FRUITS, the selection of compounds that can be stably produced by a microorganism is expanded.

Overview

Cheap-lunch strategies are those where the target compound is not directly coupled to the network of the microorganism but where a set of additional reactions is needed for its production. These reactions, however, use a directly growth coupled non-native compound as substrate. We have created a set of functions that searches for a directly growth coupled path for a substrate of a reaction producing the desired product. The functions are used recursively (i.e. using a function that calls itself) to keep searching backwards until a substrate that could be directly growth coupled was found. The functions have been tested on a toy database, containing paths for direct as well as indirect growth coupling. Ultimately, we want to test these methods when receiving information about the reactions from the databases merged by the merger and information about the network is obtained from the network transformed with methods described above. The cheap-lunch strategy-finder gets its name from the expression that ‘there is no such thing as a free lunch’, suggesting that there is always a cost attached to forming a product using multiple reactions (e.g. the energy required to produce an enzyme). The finder methods try to find the path requiring minimal energy investment for the cell: it finds the ‘cheapest’ pathways.

A new class was built within the Forbidden FRUITS API to incorporate the new functions that executed the path finding for cheap-lunch strategies. The new class can be used now as the default strategy finding class, and it takes care of searching for, not only directly growth coupled strategies but for cheap-lunch strategies as well; be it that direct growth coupling is not available for the target product or that finding a cheap-lunch strategy is possible.

Key takeaways:

Forbidden FRUITS is capable of finding a new type of strategies, which expands the selection of compounds whose production can be growth coupled to a microbial metabolic network.
Forbidden FRUITS is capable of finding a new type of strategies, which expands the selection of compounds whose production can be growth coupled to a microbial metabolic network.
The new class incorporated into Forbidden FRUITS is made to operate over the results from the merger and the network transformation tool. Its validation on a complete database is still pending.

Plumber

The plumber calculates the minimal number of modifications in an efficient and precise manner. The resulting modification set ensures that the production pathway is the only way through which biomass can be formed.

Overview

While developing the Plumber we aimed to write a flexible linear program, which connects the production of a native metabolite to the formation of biomass in such a way that this so-called coupling metabolite will become essential for biomass formation. We use a linear program consisting of the network stoichiometry, duality, indicator variables and desired/undesired phenotypes. We validated the program using toy models and the full genome scale model of Synechocystis (iSynCJ816 (Joshi, et al. 2017) and Synechocystis). Possible solutions to these models were found using Flux Balance Analysis (FBA) and Flux Variability Analysis (FVA). The program we have developed with the current set of constraints is able to correctly predict the combination of knockouts and knockins for the toy models. The program is not only able to correctly predict strategies when a production pathway is added to the native network, but also when merely a reaction to the coupling metabolite is added. This shows that the linear program is flexible to create in silico mutants which are dependent on a specific metabolite. This can be used to create a chassis, which is easily amenable to produce a variety of compounds. However, the program does not generate correct strategies on the full scale Synechocystis metabolic network yet.

Key takeaways:

We have developed a linear programming which is able to correctly predict a minimal modification strategy to add a production pathway to a metabolic network in a growth coupled manner for toy databases
Because of the general approach, the plumber can be used in the design of chassis dependent on a specific metabolite.
The program needs to be adjusted for full genome scale models

Optimization using Downstream Metabolite Information

The product yield of the genetic engineering strategies generated by the Plumber can be improved using information about metabolites downstream. By improving the product yield, we can generate chassis more relevant for the industry.

Overview

In the metabolic network of an organism in which the strategy of the Plumber is implemented, the total flux will be divided over different pathways alongside the product forming pathway. By replacing some pathways upstream of the product reaction to a location downstream of it, a higher flux can be achieved through the production reaction. This eventually leads to a higher product yield. The algorithm first identifies the essential metabolites and the downstream metabolites of the product reaction. This information is used to create a branch (path) from the essential metabolite to one of the downstream metabolites. The algorithm searches the database for reactions which can be used to extend the path from the essential metabolites to the downstream metabolites. This path consists of a concatenation of reactions. The completed path is tested for feasibility using FBA. As a result, the algorithm returns all possible ways to connect the essential metabolite to the downstream metabolites. The current program is working on a toy model. We are currently designing more toy models with more exceptions to test the program.

Key takeaways:

Product yield can be improved by redirecting flux in the network
We have written a program which can modify the network to improve product yield
We need to test the current program on more complicated toy models and on a prediction using a full genome scale model
We aim to incorporate this optimization method in the linear program of the Plumber to decrease the computational complexity.

Sequence optimization for Heterologous Expression

The sequence optimizer offers users of Forbidden FRUITS a tool to increase the success rate of gene knock ins.

Overview

Synonymous codon substitutions may sound innocent, but substitutions may not always be silent. Evolutionary selection has tuned local translation rates to improve the efficiency of co-translational protein folding [6]. Especially when genes are used for heterologous expression, the difference between synonymous codon usage of species can alter the amount of protein synthesized and the yield of correctly folded protein [7]. It seems important to introduce a component in the Forbidden FRUITS-algorithm that is able to use evolutionary differences in synonymous codon usage to optimize sequences from the original organism for heterologous expression in the host organism.

Key takeaways:

We designed a strategy to build a sequence optimization tool
Sequence optimization can increase the success rate of gene knock ins

Wetlab

Producing salicylic acid in Synechocystis PCC6803

We validate that the algorithm can predict viable strategies to generate a pyruvate dependent strain, which could also be used to make a variety of other products, by implementing the strategy to produce salicylic acid in engineered Synechocystis PCC6803 (Synechocystis).

Overview

The following modifications were predicted by Forbidden FRUITS:

Knockout: ddh, pta, acs, me, pyk1, pyk2
Knockin: irp9

A ∆acs∆pta∆me∆ddh∆pyk1 Synechocystis strain with irp9 expressed under control of the J23111 promoter was constructed at the Molecular Microbial Physiology (MMP) group at the University of Amsterdam (UvA). However, knocking out the last gene, pyk2, had not yet been successful at the beginning of our project.To ensure that enough pyruvate is formed by the production pathway, we tried to add an additional copy of irp9 under the expression of the strong BBa_J23119 promoter at the pyk1 genomic locus. However, Synechocystis was only able to incorporate this construct when irp9 is mutated. Since we were not able to obtain a gene construct with irp9 under expression of the the BBa_J23119 promoter in E.coli as well, we hypothesize that high expression of irp9 might disrupt the metabolic network of the cell, and therefore, only Synechocystis and E.coli cells with a dysfunctional irp9 gene are able to survive. Secondly, there are suggestions that the gene neighbouring pyk2, sll1276, is an essential gene (Yao, et al. 2020). Therefore, we introduced a promoter between pyk2 and sll1276 before knocking pyk2 out using CRISPR-Cpf1. Although we did obtain colonies after performing CRISPR-Cpf1, and it is looking very promising, results still need to be validated, which unfortunately is not possible within the timeframe of iGEM2020.

Key takeaways:

High expression of irp9 might disrupt metabolism of both E.coli and Synechocystis
Two new strains were tentatively constructed using CRISPR-Cpf1: ∆acs∆pta∆me∆ddh∆pyk1 with irp9 expressed by the J23111 promoter and with sll1276 expressed by the BBa_J23104 promoter and a similar strain with sll1276 expressed by the BBa_J23119 promoter.
Correct introduction of the BBa_J23104 and BBa_J23119 promoters and the ability to knock out pyk2 needs to be confirmed.

Producing salicylic acid in E. coli K-12

In this part of the project, we demonstrate the implementation of a growth-coupled strategy to produce salicylic acid in a pyruvate-dependent E. coli. Together with the production of salicylic acid within Synechocystis PCC6803, this will prove that the algorithm can handle different metabolic networks from a diverse range of host organisms to produce an identical compound (here, salicylic acid).

Overview

The following predictions were made by Forbidden FRUITS to achieve growth-coupled production of salicylic acid in E. coli K-12:

Knockout: pykA, pykF, maeB, gldA
Knockin: irp9

For this project we used a pyruvate auxotrophic strain from Wang and colleagues, specifically E. coli JW6 (E. coli ATCC 31884 ΔpheLA, ΔtyrA, ΔpykA, ΔpykF, Δgld, ΔmaeB and ΔtrpE). [1] In this specific study, they incorporated anthranilate synthase (trpE) which converts chorismate in anthranilate and pyruvate. Instead, we aim to incorporate salicylate synthase (irp9) which converts chorismate in salicylate and pyruvate. For this reason, a plasmid (pFL-XS) containing an antibiotic cassette (ampicillin) and promoter J23111 and gene irp9 has been constructed (pFL-irp9) through molecular cloning.

Until now, we have verified the pyruvate dependency of E. coli JW6 through cultivation in M9 minimal medium. At this moment, we are transforming the pFL-irp9 into E. coli JW6 to achieve pyruvate-coupled production of salicylic acid. The resulting strain, JV1 (E. coli JW6 with pFL-irp9), will be plated on M9 minimal medium plates with and without pyruvate to check if JV1 can produce enough pyruvate to support its own growth.

Key takeaways:

E. coli JW6 has been shown to be pyruvate dependent and could grow in M9 minimal medium supplemented with pyruvate
The introduction of pFL-irp9 in E. coli JW6 has been verified by plating on ampicillin M9 minimal medium plates, and subsequent PCR and gel electrophoresis using primers annealing to the pFL-irp9
Ongoing growth-related tests are currently conducted, and, if necessary, we will resort to adaptive evolution experiments to improve growth kinetics of the engineered salicylic acid producing strain.

Producing lactate in Synechocystis PCC6803

By applying a genetic engineering strategy predicted by Forbidden FRUITS for the growth coupled production of lactate to Synechocystis PCC6803, we show that the algorithm is able to couple the production of a non-native compound to growth engineered bacteria, using oxaloacetate as a coupling metabolite.

Overview

The following strategy was predicted by Forbidden FRUITS in order to achieve growth-coupled production of lactate in Synechocystis:

Knockout: mdh, me1 and ppc
Knockin: mlth and me2

Four mutants constructed at the Molecular Microbial Physiology (MMP) group at the University of Amsterdam (UvA) were used in our project:

smg001: ∆mdh with mlth expressed by the PcpcBA promoter
smg002: ∆mdh with mlth expressed by the PpsbA2 promoter
smg003: ∆mdh ∆me1 with mlth expressed by the PcpcBA promoter
smg004: ∆mdh ∆me1 with mlth expressed by the PpsbA2 promoter

However, cells were not viable when ppc was knocked out. We hypothesize that this is due to a low reaction rate going from phosphoenolpyruvate (PEP) to pyruvate. We aim to overcome this barrier and achieve this last knockout by conjugating a replicative plasmid containing pyk2 and the PcpcBA promoter in the strains mentioned above. As a consequence PEP would then be converted to pyruvate at a higher rate, which enables us to knockout the ppc gene. At the moment, we have performed a transformation to replace ppc by me2. We have generated 8 new strains building upon the SMG001-004 strains by inserting a DNA fragment with the me2 gene expressed by different promoters (PspbA or PspbA2). Additionally, to achieve transformation of these strains, we created and inserted a replicative plasmid with pyk2 under the expression of the PcpcBA promoter and with a spectinomycin resistance cassette. By knocking in me2, we ensure that enough L-Malate is formed. L-Malate is essential for both the production of Oxaloacetate and L-Lactate.

Key takeaways:

The reaction from phosphoenolpyruvate (PEP) to pyruvate is possibly too slow to convert all PEP produced in the Calvin Benson Cycle. This could cause a build-up of metabolites upstream of PEP when the ppc gene is knocked out.
We generated a replicative plasmid with pyk2 under the expression of the PcpcBA promoter and with a spectinomycin resistance cassette.
We created 8 new strains, which are built upon the four initial strains to perform a knockout of ppc and knockin of me2 using different strategies.

Producing lactate in Synechococcus UTEX 2973

We further illustrate how Forbidden FRUITS can predict strategies to make a variety of non-native compounds via the oxaloacetate node on the same microbial metabolic network, by constructing a lactate-producing microbial system in Synechococcus UTEX 2973.

Overview

The following strategy was predicted by Forbidden FRUITS in order to achieve growth-coupled production of lactate in Synechocystis:

Knockout: ppc
Knockin: me, mlth

In Synechococcus UTEX 2973 (UTEX), we have successfully constructed both knockin and knockout plasmids using a strategy identified by Forbidden FRUITS. The knockin plasmid was successfully created with backbone pSEVA451_oriT_F through digestion and ligation of the mlth and me genes. We are creating four varieties of this plasmid to explore which will yield the best results in UTEX, essentially using all combinations of strong and weak promoters for each of the knockin genes. In the process of creating the knock-in plasmids, we had some mutations arise in the me gene. These mutations were successfully repaired through ligation and fusion PCR using the corrected gene. To construct our knockout fragment of ppc, we used the backbone pFL-SN_mob, which now successfully contains homologous regions for ppc with a kanamycin cassette.

Key takeaways:

Successful construction of ppc knockout plasmid with a pFL_SN_mob backbone
Successful engineering of four knockin plasmids with pSEVA451_oriTOF: all combinations of the mlth and me genes with strong and weak promoters.
We are currenting in the process of conjugating these plasmids in UTEX and are waiting for full segregation to explore how well this strategy works at scale.

Producing mannitol in Synechocystis PCC6803

In order to show that the cheap lunch strategy finder predicts feasible strategies, we applied one of these strategies to Synechocystis PCC6803. In this strategy, mannitol is produced in a growth coupled way, as mannitol production is not directly incorporated back in the metabolic network.

Overview

The following strategy was predicted by Forbidden FRUITS in order to achieve growth-coupled production of mannitol in Synechocystis PCC6803 (Synechocystis):

Knockout: glgc
Knockin: susA, mdh

To incorporate this strategy, we used a strain from the Molecular Microbial Physiology (MMP) group at the University of Amsterdam (UvA) where the mdh gene was already incorporated. We tried to replace the glgC gene by susA by making a vector that could introduce susA, flanked by two antibiotic resistance genes (Omega and CmR), using a double crossover on the glgc gene. Next, the cells were grown in the presence of antibiotics to push segregation (replacing all copies of glgc by susA). However, even in high antibiotic concentrations segregation was not successful. Even after adding sucrose to the medium, which the cells need for the reaction catalyzed by susA, we did not manage to obtain a fully-segregated isolate. Since mannitol has been shown to be a solute in osmotic stress conditions (Chaturvedi et al., 1997), we tried to increase the osmotic stress of the cells by increasing the NaCl concentration in the medium. Cells producing higher concentrations of mannitol would also be more stress resistant, making mannitol production more attractive for the microbes. Next to this, we set up a line of microbes that would also grow without light, as this would make the glycogen production not essential since this energy storage is used during nighttime (and thus not produced). In order to accommodate the lack of energy in the dark condition, we added glucose to the media for the bacteria to grow.

Key takeaways:

Successful knockin of susA into the genome.
Despite all efforts to push full segregation of susA, we are still in the process of completing this step.

References

[1]Wang J, Zhang R, Zhang Y, Yang Y, Lin Y, Yan Y. Developing a pyruvate-driven metabolic scenario for growth-coupled microbial production. Metab Eng. 2019;55:191–200.

[2]Joshi C. J., Peebles C. A. M., Prasad A. (2017). Modeling and analysis of flux distribution and bioproduct formation in Synechocystis sp. PCC 6803 using a new genome-scale metabolic reconstruction. Algal Res. 27, 295–310

[3]Yao, L., Shabestary, K., Björk, S.M. et al. Pooled CRISPRi screening of the cyanobacterium Synechocystis sp PCC 6803 for enhanced industrial phenotypes. Nat Commun 11, 1666 (2020).

[4]Machado D, Herrgård MJ, Rocha I. Stoichiometric Representation of Gene–Protein–Reaction Associations Leverages Constraint-Based Analysis from Reaction to Gene-Level Phenotype Prediction. Patil KR, editor. PLOS Comput Biol. 2016;12: e1005140.

[5]Chaturvedi V, Bartiss ANN, Wong B. Expression of bacterial mtlD in Saccharomyces cerevisiae results in mannitol synthesis and protects a glycerol-defective mutant from high-salt and oxidative stress. J Bacteriol. 1997;179(1):157–162.

[6]Jacobs WM, Shakhnovich EI. Evidence of evolutionary selection for cotranslational folding. Proc Natl Acad Sci U S A. 2017;114(43):11434–9.

[7]Rodriguez A, Wright G, Emrich S, Clark PL. %MinMax: A versatile tool for calculating and comparing synonymous codon usage and its impact on protein folding. Protein Sci. 2018;27(1):356–62.