Design
The world is shaped by interactions between living things. Although these relationships represent one of the foundations for diversity and complexity of life, they remain largely unexplored. To investigate these networks, researchers are trying to mimic these interactions with co-cultivation.
But we do not want to mimic, we want to understand.
We are convinced that we can only get a deeper understanding of these associations if we are able to build them from bottom-up. Our modular co-cultivation toolbox is intended to contribute to this. Every scientist should be able to design and use co-cultures with different strains or organisms for diverse purposes with the help of our toolbox. But what is the basic science behind it and how does this work?
Scientific Background
The Root Problem
Only 1% of microorganisms have been successfully cultivated so far1, whereas the remaining 99% require conditions that are not yet reproducible in monocultures or with standard cultivation methods.
Microorganisms in nature are mostly, if not always, found in microbial consortia such as in the human gut or in the soil. Some microorganisms may struggle to grow in monocultures due to the lack of essential benefits arising through the coexistence with other microbes that can include cross-feeding and signaling molecules, among others2.
A problem for science is the complexity of networks between microorganisms in their natural environment. Studying these naturally occurring communities is very difficult since the establishment of such a co-culture is a big challenge by itself. For every microorganism, different needs need to be met such as media composition, pH or temperature and survival of all organisms in the co-culture needs to be ensured.
Thus, a lot of discoveries (e.g. production of biofuels and medical applications) are still neglected. This includes research about the gut microbiome, which has an overall effect on the human health. And while their cooperative interactions are not fully understood until now, the effect of drugs on every human being with differing gut bacteria cannot be calculated yet. Measures for keeping a healthy gut microbiome often remain ineffective since they are tested on monocultures.
Our Approach
The co-cultivation of microorganisms has multiple useful applications from basic research for cell-cell interactions to medical treatments for bacterial infections or industrial production of bioproducts. A simplified modular approach to co-culturing offers potential benefits for every aspect of our society.
In recent years, synthetic biology picked up the problem to co-cultivate microorganisms through synthetically created dependencies. We tried to tackle this challenge by creating not only three different dependencies, but also for three different organisms. The aim of our project was to create a modular toolbox for co-culturing of Escherichia coli, Saccharomyces cerevisiae and Synechococcus elongatus as a big advancement towards possible co-culturing between all known organisms.
With these organisms our team engineered a system based on nutrient exchange. Here phosphate is provided through oxidation of phosphite, nitrogen source produced by melamine breakdown, whilst a carbon source is provided by
To make our co-culture toolbox as modular as possible, we used Golden Gate cloning3,4 to design constructs with exchangeable parts in order to easily customize them to the needed co-culture. Therefore, we decided to build our systems with already established toolboxes: the YTK toolbox5 established in John Dueber's laboratory for
The advantage of Golden Gate cloning is the usage of type II S restriction enzymes that do not cleave directly at the recognition site, but at a specific number of base pairs up - or downstream. This allows the design of scarless constructs and specific overhangs. As a result, different parts are assembled only in the correct order. In consequence, one-pot reactions with several parts are possible.
The YTK (Yeast toolkit) established by the Duber laboratory5 is built up in different levels.
All backbones are cassettes that contain different antibiotic resistances, origin of replication and drop-outs. The backbones are chosen that best fit the cloning.
The level 0 backbone serves as saving and transporting the inserted gene. The gene of interest is cloned into the so called entry vector pYTK001 using the enzyme BsmBI. For this to work, recognition sites of BsmBI have to flank the gene and the overhangs must be designed to fit to overhangs of the restriction sites of the entry vector. The entry plasmid contains a chloramphenicol resistance and an origin of replication but no promoter or terminator so there is no expression of the insert.
The level 1 vector follows for cloning an expression plasmid for yeast. It always includes an appropriate promoter and terminator for the gene to be ready to be cloned into S. cerevisiae. Therefore, a promoter, a gene of interest and a terminator are chosen from the toolbox, where over 20 promoters - constitutive and inducible - and different terminators are selectable. In contrast to the level 0 plasmid, the level 1 vectors contains an ampicillin resistance and recognition sites for the restriction enzyme BsaI.
With a GFP dropout system the cloning is facilitated. A gfp gene is located between the restriction sites in the backbone and will be replaced if the gene of interest is correctly inserted. In case of non-recombinant bacteria due to wrong ligation or failed cloning, the colonies show a green color.
In case it is required to express more than one gene in a yeast strain, a level 2 plasmid is available as well which can contain up to three transcriptional units.
For E. coli, the CIDAR toolbox of Iverson and colleagues is used for modular cloning with Golden Gate6. The concept is analogous to the YTK.
To generate a level 0 plasmid, the gene of interest is flanked by BbsI recognition sites with fitting overhangs and is cloned into an entry vector with ampicillin resistance.
The level 1 plasmid is a functional E. coli expression plasmid. For this vector, ten different constitutive or controllable promoters, six different ribosome binding sites (RBS) and a terminator can be found in the toolbox. The plasmid contains a kanamycin resistance and recognition sites for the enzyme BsaI.
In the CIDAR toolbox a multi-gene assembly plasmid with a maximum of four transcriptional units is possible as well in the level 2 plasmid. Just like the level 1 plasmid, the encoding resistance is ampicillin and the enzyme BbsI is used for restriction.
For all cloning steps, blue white screening can be used for the identification of recombinant bacteria. Therefore, the backbone contains lacZ which encodes ß-galactosidase with an IPTG inducible promoter at the insert point to help find positive colonies. ß-galactosidase hydrolyzes X-Gal into a molecule which spontaneously oxidizes into a blue pigment. Thus, agar plates with IPTG and X-Gal are used7.
After incubation, the colonies show either a white or a blue color. Blue colonies only contain the backbone whereas white colonies carry the gene of interest instead of the functional lacZ gene.
Design of our Sub-projects
Self-lysing Microorganisms
Growing cultures containing more than one organism without one or more of the populations dying out is a very complex task. Since our aim is not only to grow a co-culture of two microorganisms but of three different subjects, we decided to develop a tool to downregulate an organism's growth in a controlled way as a first step towards a stable co-culture. This part is essential because it permits the co-existence of at least two organisms at the same time, without erasing one another due to growth disadvantages.
For this purpose we designed two mechanisms that are supposed to downregulate the growth of the two fastest-growing organisms in our final co-culture: E. coli and S. cerevisiae.
Our models for the two different organisms are based on the same principle: Quorum sensing. This mechanism is used by many organisms to regulate gene expression depending on cell density fluctuation. This machinery is often referred to as the communication system of bacteria and is based on the activity of so-called autoinducers, such as acyl homoserine lactones (AHLs)8. The most commonly known quorum sensing regulatory mechanism is from the bacterium Vibrio fischeri. Depending on the population density, the amount of autoinducers fluctuates and, when elevated, can alter gene expression on a global level - in the case of V. fischeri, by activating bioluminescence9. To apply this mechanism, we inserted a lysis gene under the control of the AHL-stimulated Lux promoter, meaning that when the quorum sensing molecule, acyl homoserine lactone synthase is produced by the luxI gene and then binds to the regulator LuxR, expression of the lysis gene is started. At high cell densities, this promoter is activated and the cell population is thereby simply regulated by itself.
Experimental Design
All our constructs are virtually cloned and adapted to the CIDAR and YTK toolboxes that we are using. For assembly with the toolbox, each part has to be cloned first into so called level 1 plasmids. Those are then brought together to the final functional system. The use of those toolboxes does not only allow the diverse assembly of various parts, but also the exchange of different promoters, RBSs and terminators, with individual strength and preferences.
As an example, all level 0 virtual constructs, as well as all level 1 virtual constructs of our Lux quorum sensing system are shown in Fig. 1. Essentially, the level 0 constructs represent the basic part in one of the toolbox backbones.
The level 1 assemblies should be functional units, comprising promoter, RBS, gene of interest and terminator.
In our case, only the level 2 constructs are functional, since the quorum sensing system works only if all parts are assembled together.
We codon optimized the Lux System for both organisms and put the lysis gene E from bacteriophage phiX174 under the control of Plux in E. coli10. This lysis protein induces host cell lysis by inhibiting the host cell’s translocase MraY activity, which is required for the synthesis of an important cell wall component. Lack of function of this protein prevents the bacterial cell from growing properly and as a result leads to final lysis. We constructed a plasmid, containing all genes necessary for the activation of this promoter: luxI and luxR. To measure the cell population using fluorescence our construct contains a gfp reporter as well, therefore the activation of the lysis gene leads to the co-activation of the reporter to keep track of the cells.
Besides this final construct, we also created different controls. This includes a plasmid containing all the genes except the luxI, which is the acyl homoserine lactone synthase. It was created to show that the promoter is activated and lysis or reporter expression is functional only upon artificial induction with AHL. Moreover, we also designed a plasmid which is supposed to show functionality of the Plux promoter, hence exchanging the lysis gene E with gfp.
Overproduction of Leucine by S. cerevisiae
For yeast we implement a leucine producing, but lysine-auxotrophic yeast strain.
Our initial strain is S. cerevisiae BY4742 (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0). It is a deletion strain with auxotrophies for leucine, lysine, histidine and uracil. The strain is normally used for work with numerous non-antibiotic selection markers. Important for our use is the null mutation leu2Δ0. LEU2 is an essential gene in the leucine biosynthesis and is coding for the beta-isopropylmalate dehydrogenase (IMDH). IMDH catalyzes the third step in the leucine biosynthesis pathway.
Experimental Design
We are not only restoring its ability to produce leucine but are improving its production: To repair the leucine auxotrophy, we cloned LEU2 with a strong constitutive promoter into the yeast cells. This way, we added the missing gene which caused the auxotrophy. With the addition of the mentioned promoter, we are overexpressing LEU2.
Through its originally derived auxotrophies, the yeast is lysine-auxotrophic and fits perfectly for our design.
Overproduction of Lysine in E. coli
To create an artificial dependency between E. coli and S. cerevisiae based on amino acid auxotrophies we want E. coli to produce and secrete lysine for the lysine auxotrophic yeast.
At the same time E. coli is auxotroph for the amino acid leucine. To achieve overproduction of lysine we want to make use of the protein meso-diaminopimelate D-dehydrogenase (Ddh) encoded by the gene ddh11 and a non feedback inhibited version of the enzyme aspartokinase (LysC) encoded by the gene lysC12. The native organism of both genes is Corynebacterium glutamicum.
ddh is a multistep enzyme which catalyses the reductive amination from L-A'-piperideine-2,6-dicarboxylate to D,L-2,6-diaminopimelate. This leads to an alternative pathway for the lysine biosynthesis in C. glutamicum which we want to implement in E. coli.
Aspartokinase taken from the wild type E. coli is feedback inhibited by lysine. An amino acid switch from serine to phenylalanine at position 381 results in a non feedback inhibited version of LysC which leads to an improved lysine production.
Experimental Design
The gene for lysC was ordered codon-optimized as a gBlock, whereas the gene ddh was isolated from C. glutamicum using the Quick-DNA Miniprep Plus Kit.
The cloning was performed according to the CIDAR MoClo toolbox12.
Since the gene ddh contains recognition sites for the two restriction enzymes BsaI and BbsI, that we use in the CIDAR MoClo toolbox cloning, we had to perform two site-directed mutagenesis. Therefore, the gene ddh was first stored in the TOPO cloning vector using TA cloning and afterwards mutagenized two times.
We designed six different expression plasmids for each of both genes. They differ only in the constitutive promoters we used in order decide which promoter strength is needed for the amino acid overproduction needed for the auxotrophic 
To create the expression plasmids, the level 1 backbones, the level 0 plasmid with the gene of interest as well as promoter, ribosome binding site and the terminator are cut with BbsI and afterwards ligated with the level 1 backbone vector.
Overproduction of Sucrose by S. elongatus
A carbon source is needed in our co-culture system for E. coli and S. cerevisiae in order to stabilize the culture and to create an autonomous system. This is why we included the cyanobacterium Synechococcus elongatus PCC 7942 in our three-way co-culture since it is a photosynthetically active cyanobacterium. By engineering S. elongatus to overproduce and secrete sucrose, we aim to create a stable co-culture that is not further dependent on carbon in the media. S. cerevisiae is able to convert sucrose into fructose and glucose extracellularly which can then be used as a carbon source by the S. cerevisiae itself13 or the E. coli.
Many cyanobacteria are able to overproduce substances like sugars in order to prevent the cell from osmotic stress14.
In addition it could be shown that upon induction of the symporter CscB, S. elongatus is able to produce higher amounts of sugar. S. elongatus produces sucrose in several steps from Glucose-1-P. The first rate limiting step in the pathway for the sucrose production is catalyzed via sucrose-phosphate synthase (Sps) and is induced by salt stress15.
We obtained a specific strain published by Dr. Daniel Ducat et al. which is able to secrete sucrose with the help of the CscB sucrose permease. The cscB gene encodes a symporter which is inducible by IPTG13. In this case we plan to induce sucrose production with NaCl and IPTG. Furthermore, advice for cultivation media for the co-culturing experiments were communicated by Daniel Ducat. This is also the foundation for the Integrated Human Practice.
Our goal is to establish a system which allows the production and the secretion of sucrose via the induction with only IPTG instead of NaCl. With this approach we want to avoid salt stress for the culture.
Experimental Design
We are cloning an expression plasmid which contains an iGEM provided IPTG inducible promoter (BBa_R0010, termed Plac here) and sps (sucrose phosphate synthase) as the gene of interest. The origin species of sps we used is Synechocystis PCC 680313. This variant of sucrose phosphate synthase needs less NaCl in order to be activated, compared to the variant already found in S. elongatus14.
We want to insert the plasmid into our strain of S. elongatus by conjugation. We hypothesize that the induction with IPTG is enough to induce the production and the secretion of sucrose.
For the cloning strategy of our expression system, Gibson Assembly was chosen. The used backbone is called pSHDY, which was created and provided by an advisor of our team. pSHDY is a conjugative shuttle vector with a broad host range which is commonly used as a plasmid-based expression system in cyanobacteria. Selection markers are antibiotic resistances - in this case kanamycin and spectinomycin. The strain we are working with provides a genomically integrated chloramphenicol resistance which also allows the specific selection of positively conjugated clones. For the uptake of plasmid DNA, the method triparental mating17 was used.
Production of Phosphate from Phosphite by S. cerevisiae
For our nutrient based dependency system we engineered S. cerevisiae to provide phosphate as a phosphorus source for the other two organisms.
A recent approach by Kanda et al18 for a new selective marker in yeast is based on the oxidation of phosphite (PT) to metabolically useable phosphate (PI). The used phosphonate dehydrogenase encoded by ptxD caught our interest in using phosphite as a phosphorus source for our co-culture.
Experimental Design
Based on the work of Kanda and colleagues a codon-optimized version of ptxD for S. cerevisiae was ordered from IDT. Additionally Golden Gate assembly overhangs based on the design by Lee et al19 in the pYTK MoClo Golden Gate assembly toolbox a type 3 CDS were added as well as biobrick assembly prefix and suffix at the 3` and 5` end of the sequence.
Concordant to the pYTK cloning workflow the optimized ptxD (ptxD_opt) was inserted into a level 0 entry vector
These level 0 vectors were then used to create yeast expression vectors.
The expression plasmids always carried the ptxD_opt and the tPGK1 terminator sequence as a fixed part of the newly assembled transcription units. For the promoter sequence five different constitutive yeast promoters were used to create 5 different expression plasmids respectively. The five used promoter regions are pTDH3, pCCW12, pPGK1, pHHF2 and pTEF1.
Production of Ammonia from Melamine by E. coli
In order to create an artificial dependency for our three-way co-culture based on cross-feeding of essential nutrients, we developed the idea to engineer E. coli to supply the other two organisms with ammonia.
While researching for papers and taking a look at several already established ammonia pathways we came across an idea published by Shaw and colleagues in 2016 who suggested a pathway cluster made of six enzymes from different organisms for the conversion of melamin into ammonia20. We implemented this melamine pathway cluster in E. coli according to Shaw and colleagues in order to reduce one mol melamine to six mol of ammonia in only six conversion steps.
The cluster contains three enzymes performing hydrolase reactions following another three deaminase reaction enzymes.
It consists of a melamine deaminase (triA) from Acidovorax avenae, an ammeline deaminase (guaD) from E. coli and an ammelide deaminase (trzC) from A. avenae. The hydrolase reactions are executed by a cyanuric acid hydrolase (atzD) from Pseudomonas sp., a biuret hydrolase (trzE) from Rhodococcus sp. and an allophanate hydrolase (DUR1,2) as found in 
Since ammonia diffuses through the membrane of E. coli on its own, the design of an exporter was not needed.
Instead of cloning the genes from the different, native organisms, the codon-optimized genes were ordered as gBlocks - except for guaD which was directly amplified from the genome of E. coli DH5α.
Due to the staggering amount of genes, the inserts were divided on to two plasmid vectors. The genes of the first four reaction enzymes (guaD, trzC, atzD, triA) as well as promoters, ribosome binding sites and terminators were equipped with different overhangs depending on their position in the final plasmid and were cloned separately into the level 0 vectors of the CIDAR toolbox.
In the following step, each level 0 plasmid was cut with the enzyme BsaI. The gene of interest was inserted into the level 1 plasmids with its corresponding promoter, RBS and terminator. Due to the different overhangs of each part they assembled correctly.
Next, the level 1 plasmids of the four genes were cut with the enzyme BbsI to retrieve the genes of interest with corresponding promoter, RBS and terminator. Due to the varying overhangs on the promoter and terminator sequence they bind to one another, resulting in the final level 2 plasmid for transformation in 
The other two genes trzE and DUR1,2 were cloned into a second plasmid via Gibson Assembly to evaluate a different cloning method. It is a molecular cloning method that allows the assembly of of multiple DNA fragments in a single reaction with the help of overlapping overhangs and exonucleases that cleave DNA at the 5’ end.
Since the DUR1,2 gene was too long for one synthesis, it had to be ordered in two parts. Following an overlap PCR of trzE and the first part of DUR1,2 the two inserts were assembled into pKIKO, a plasmid for genomic integration in E. coli.
For our project, an understanding of the growth of the three different organisms
Therefore, we observed the growth behavior of these three organisms in different media compositions.
Material and Methods
During the experiments, the flasks - containing the monocultures in 20% of total flask volume (e.g.
To keep track of growth, the optical density (OD) of each culture was determined with a spectrophotometer.
Because of differing division rates among the three organisms, the measurements were performed over a varying period of time with individual measurement intervals adapted to each organisms generation time.
S. cerevisiae and E. coli were observed over a timeframe of 36 hours, whereas the growth of S. elongatus was observed over the course of a week. The OD of S. cerevisiae was measured every 2 hours, E. coli every 0.5 hours and S. elongatus every 12 hours.
For our experiments the organisms were inoculated in different media to test various cultivation conditions:
Lysogeny broth (LB) is the standard cultivation media for E. coli and contains different minerals, essential amino acids, organic compounds, vitamins and certain trace elements.
Yeast extract peptone dextrose (YPD) is a complete media for S. cerevisiae and includes yeast extract, glucose and amino acids.
BG11 - also called “blue green medium” - is the standard cultivation media for blue green algae such as cyanobacteria. It does not include a carbon source since the photoautotrophic organisms produce it themselves.
M2 medium is a buffered derivate of BG11 with a higher amount of Mg2+, additional phosphate and less nitrate21. It is our three-way co-culture medium suggested by Daniel Ducat.
Synthetic defined media (SD) is a minimal medium for S. cerevisiae. It can be used with synthetic dropout supplements for auxotrophic yeast strains.
Results
Growth of E. coli
E. coli has a relatively high growth rate (0.5274 h-1) in the first four hours in its commonly used cultivation medium LB. It reaches a maximum OD600 of around 6 after 10 hours and stays in the stationary phase for the rest of the measurement. In comparison, growth was fairly reduced in M2 medium, lacking essential carbon sources. The optical density increased very slowly, but constantly with a growth rate of 0.0326 h-1, never reaching an OD600 of 1.
Growth of E. coli in BG11 with additional Sugar and Salt stress
As our three-way co-culture requires salt in order to induce the sucrose secretion of
Growth of S. cerevisiae BY4742
S. cerevisiae was observed in all media. It grows well in YPD (0.692 h-1) and SD media including amino acids (0.6719 h-1) with an OD600 of around 10 after 36 hours. It also shows growth in LB (0.518 h-1). It doesn’t grow as well in the sugar-deprived M2 medium with the OD600 decreasing after an initial growth spurt (0.4904 h-1), and barely shows any growth in BG11 with a growth rate of 0.0165 h-1.
Growth of S. elongatus
S. elongatus grows relatively slow compared to the other two organisms with a generation time of 12 hours.
There is a slight difference between the growth of S. elongatus in M2 medium and BG11 with growth rates around
However, no growth occurred in YPD. Instead, the monocultures show a decreasing OD600 over time, which indicates that the cells were dying. Therefore, the measurements were terminated after 4.5 days.
For our three-way co-culture, we contemplated using the
Discussion
With our growth curve measurements we were able to make sure that consistent growth patterns exist for all of our three organisms.
We predicted that E. coli would overgrow the co-culture over time since its generation time is usually around 0.5 hours. It is is a lot shorter compared to the other two organisms in the planned three way co-culture, which would complicate the establishment of a co-culture with E. coli. This can be observed during the growth development in LB. After a short exponential phase of 4 hours with a growth rate of 0.5274 h-1, E. coli cultures enter the stationary phase, having reached the maximum OD600 of around 6 (Figure 1).
However, OD600 in the M2 medium increases at a slower rate, but constantly, thus making the medium optimal for a co-culture including E. coli.
S. cerevisiae grows very well in YPD, SD and LB media and shows growth in M2 medium to some extend before OD600 is decreasing. There is almost no visible growth in BG11 (Figure 3). We assume that the initial growth spurt of
In addition to that, we also proved that S. elongatus
Based on the information we gained with the experiments performed on the wild type strain, we assume similar behavior for the cscB strain and therefore did not conduct further growth measurements in different media with the latter strain. Moreover, this medium has been used in previous, published experiments, which emphasizes our rational use of this media21.
Besides, we were able to observe that the conditions needed for inducing the cscB strain would not interfere with the growth of E. coli (Figure 2). BG11 as well as M2 medium do not contain a carbon source, which is why E. coli without any additional sucrose stops growing after one division. The growth rate is around 0.0142 h-1. We hypothesize that residual minimal growth at later time points might result from nutrients provided by dead cells in the media.
This is different with the sugar enriched media. While the growth is still reduced in the beginning, the cultures enter an exponential phase after around 20 hours. The log phase with sugar and salt is 0.1961 h-1, without salt it is 0.1645 h-1. The delay could be due to the preferred carbon source, which is glucose. It is possible that
Furthermore, we showed that M2 medium promotes the growth of
An improved experiment would be obtained with an addition of technical replicates and more biological replicates in order to reduce the standard deviation errors. In addition to that, the choice of only one incubator for the entire experiments and the usage of same sized flasks would ensure more similar conditions for every monoculture.
Furthermore, washing the cells before inoculating them in their tested media could prevent falsifying the results during the first hours of the measurement, as an induction of growth due to transferred essential nutrients from the pre culture medium is possible.
- Katz, Micah, Bradley M. Hover, and Sean F. Brady. "Culture-independent discovery of natural products from soil metagenomes." Journal of industrial microbiology & biotechnology 43.2-3 (2016): 129-141
- Sonia R. Vartoukian, Richard M. Palmer, William G. Wade “Strategies for culture of ‘unculturable’ bacteria” FEMS Microbiology Letters, Volume 309, Issue 1, 1 August 2010, 1–7
- Engler, Carola, et al. "Golden gate shuffling: a one-pot DNA shuffling method based on type IIs restriction enzymes." PloS one 4.5 (2009): e5553
- Engler, Carola, Romy Kandzia, and Sylvestre Marillonnet. "A one pot, one step, precision cloning method with high throughput capability." PloS one 3.11 (2008): e3647.
- Lee, Michael E., et al. "A highly characterized yeast toolkit for modular, multipart assembly." ACS synthetic biology 4.9 (2015): 975-986
- Iverson, Sonya V., et al. "CIDAR MoClo: improved MoClo assembly standard and new E. coli part library enable rapid combinatorial design for synthetic and traditional biology." ACS synthetic biology 5.1 (2015): 99-103
- Padmanabhan, Sriram, Sampali Banerjee, and Naganath Mandi. "Screening of bacterial recombinants: strategies and preventing false positives." Molecular Cloning-Selected Applications in Medicine and Biology. InTech, 2011.
- Miller, Melissa B., and Bonnie L. Bassler. "Quorum sensing in bacteria." Annual Reviews in Microbiology 55.1 (2001): 165-199.
- Egland, KA.; Greenberg,E.P. “Quorum sensing in Vibrio fischeri: elements of the luxI promoter”. Molecular microbiology. 1999; Volume 21, Issue 4, 1197-1204
- Scott, Spencer R., et al. "A stabilized microbial ecosystem of self-limiting bacteria using synthetic quorum-regulated lysis." Nature microbiology 2.8 (2017): 17083.
- Schrumpf, Barbel, et al. "A functionally split pathway for lysine synthesis in Corynebacterium glutamicium." Journal of bacteriology 173.14 (1991): 4510-4516.
- Bärbel Schrumpf, Lothar Eggeling, and Hermann Sahm “Isolation and prominent characteristics of an L-lysine hyperproducing strain of Corynebacterium glutamicum” Microbiol Biotechnol (1992) 37:566-571
- Lagunas, Rosario. "Sugar transport in Saccharomyces cerevisiae." FEMS Microbiology Letters 104.3-4 (1993): 229-242
- Klähn, Stephan, and Martin Hagemann. "Compatible solute biosynthesis in cyanobacteria." Environmental microbiology 13.3 (2011): 551-562
- Qiao, Cuncun, et al. "Effects of reduced and enhanced glycogen pools on salt-induced sucrose production in a sucrose-secreting strain of Synechococcus elongatus PCC 7942." Applied and environmental microbiology 84.2 (2018): e02023-17.
- Zinchenko, V. V., et al. "Vectors for the complementation analysis of cyanobacterial mutants." RUSSIAN JOURNAL OF GENETICS C/C OF GENETIKA 35 (1999): 228-232
- Kanda, Keisuke, et al. "Application of a phosphite dehydrogenase gene as a novel dominant selection marker for yeasts." Journal of biotechnology 182 (2014): 68-73
- Lee, M. E., DeLoache, W. C., Cervantes, B., & Dueber, J. E. (2015). A highly characterized yeast toolkit for modular, multipart assembly. ACS synthetic biology, 4(9), 975-986
- A. Joe Shaw,* Felix H. Lam et al. “Metabolic engineering of microbial competitive advantage for industrial fermentation processes. “ Science 05 Aug 2016 : 583-586
- Yamanaka, G., A. N. Glazer, and R. C. Williams. "Cyanobacterial phycobilisomes. Characterization of the phycobilisomes of Synechococcus sp. 6301." Journal of Biological Chemistry 253.22 (1978): 8303-8310.
- Weiss, Taylor L., Eric J. Young, and Daniel C. Ducat. "A synthetic, light-driven consortium of cyanobacteria and heterotrophic bacteria enables stable polyhydroxybutyrate production." Metabolic engineering 44 (2017): 236-245.