Different approaches to create artificial endosymbiosis have been described previously. The used methods vary depending on the host organism, including microinjection, invasion or phagocytosis for mammalian cells and a polyethylene glycol (PEG) -mediated fusion of vesicle enclosed microorganisms with yeast spheroplasts [1][2]. Even though the uptake was successful, these methods have not been utilized to date.

In order to realize our idea of a mostly self-sustainable production we had to decide for one an applicable uptake mechanism of the symbiont into the host organism. Using mainly S. cerevisiae and E. coli, we chose the aforementioned PEG-induced uptake due to its easy applicability for different yeast strains without further engineering.

The procedure, which has been described before [2][3], has been adapted and optimized for regeneration of the S. cerevisiae cell wall in liquid media. After initial growth of S. cerevisiae host cells and desired endosymbionts to a mid-logarithmic culture, the cell wall of S. cerevisiae is digested using Zymolyase, an enzyme which cleaves the 1,3-ß-glucane bonds of the yeast’s cell wall [4]. Therefore, the yeast cells are only enclosed by their lipid membrane. Additionally, this step can be universally applied to most yeast strains since they do not vary much in cell wall components rather than in the ratio of these. The digestion of the cell wall is a necessary requirement for the next crucial step: enclosure of the chosen endosymbiont by PEG liposomes.

If malonate, which is essential for yeast viability and as it serves as fuel for the biosynthesis reaction is removed from the media it has to be delivered by the invading E. coli strain. To achieve this major changes in the beta-alanine route of E. coli had to be made to direct flux towards malonic acid production [3]. We designed an operon-plasmid consisting of six different genes from various organisms including genes from E. coli, which was overexpressed to increase the yield of produced malonic acid (Fig. 3). As backbone we used the part of the pACYC184 plasmid that consists of the E. coli origin p15A ori and the chloramphenicol resistance gene under its native cat promoter as the marker-gene. The operon itself was under the control of the constitutive promoter BBa_J23108 and consists of the ppc gene from E. coli that encodes for a phosphoenolpyruvate carboxylase that catalyzes the addition of bicarbonate to phosphoenolpyruvate (PEP) to form oxaloacetate [6], the yneI gene from E. coli that encodes for a succinic semialdehyde dehydrogenase that should catalyze the oxidation of malonic semialdehyde to malonic acid [3] and the aspA gene from E. coli that encodes for an aspartate ammonia lyase that catalyzes the reaction of fumaric acid to aspartic acid [7]. Additionally, we integrated the panD gene from C. glutamicum, which encodes for an aspartate-α-decarboxylase that catalyzes the reaction of aspartic acid to β-alanine [3] and the pa0132 gene from P. aeroginosa that encodes for a β-alanine pyruvate transaminase that catalyzes the reaction of β-alanine to malonic semialdehyde [3]. Last but not least, we integrated the mae1 gene from S. pombe that encodes for a permease for malate and other C4 dicarboxylic acids [4], that should enable the strain to segregate malonic acid [5]. The first three genes of the operon (panD, aspA and pa0132) were regulated by the RBS BBa_J61101 with an related strengths of 22.7 %. The last three genes (ppc, yneI and mae1) were under the control of the RBS BBa_B0032 with a related strengths of 33.96 %. As terminator we used the double terminator BBa_B0015 with a forward efficiency of 0.984.

Both, the operon plasmid and the expression plasmid were assembled via Gibson assembly. Ppc, aspA and yneI were amplified using genomic DNA from the E. coli strain MG1655. PanD, matB and pa0132 were amplified using synthesized nucleotide sequences as templates Mae1 was amplified using genomic S. pombe DNA as template.

A different approach, to achieve the goal of reliably interacting organisms, was to set a protein-based dependency. In this scenario the yeast has an essential protein-coding gene knocked out, which then is complemented by E. coli expressing and secreting the protein. The targeted genes in yeast can vary from classic knockout genes such as trp, which can be supplemented from the medium to more unique knockouts; for example a subunit of the yeasts ribosome. The expression of the corresponding protein in E. coli should be regulated by an inducible promoter where the inducer has to be able to pass the yeast cell wall and membrane, such as T7 or lac. Additionally, this has the advantage that it can be easily controlled whether a survival of both cells is a result of the dependency or occurred for different reasons e.g. leaving out the induction.

For the export we chose two different approaches. The first, was to construct a fusion protein consisting out of the protein of interest and YebF fused to its N-terminal end. YebF occurs in the genome of E. coli and has shown to be capable of exporting fused proteins due to unknown mechanisms [8]. The second approach was similar, yet a little bit more complicated. We fused the 178 bp signal sequence of the Flagellin encoding gene fliC upstream of the coding region of the desired protein. This sequence acts as a signal for the directed transport and assembly into the flagellum. A knockout of FliC and FliD interferes with the regulation of both transport and assembly. Hence, instead of Flagellin our protein of interest is transported directly to the membrane, where it gets secreted into the medium due to defective assembly. In addition, the FliC signal sequence has shown to be cleaved during this procedure [9].