Our ternary microbial symbiosis system is consist of three chassis: Escherichia coli, Synechococcus elongatus , and Azotobacter caulinodans. S. elongatus fix CO₂ through photosynthesis and secretes the product, sucrose as the carbon source for heterotrophic chassis. A. caulinodans convert the N₂ in Mars atmosphere into the universal nitrogen source, NH₄⁺. E. coli plays the role of production chassis. We performed a series of engineering iterations to run sustainably on Mars. First, we have carefully considered engineering on S. elongatus to maximize carbon source input. Second, we optimized nutrient response and nitrogen fixation regulation part to allow our intercellular gene circuits to more efficiently support the ternary symbiotic system. Finally, we present our optimization process on the medium composition and verify that all three strains can grow in CoBG11.

As the only autotrophic organism in our ternary microbial symbiosis system, S. elongatus is expected to act as the direct carbon source supplier of E. coli and the indirect supplier of A. caulinodans in our design. Based on the following two aspects, wild-type S. elongatus cannot meet our requirements. On the one side, wild-type S. elongatus cannot secrete sucrose as a carbon source under normal conditions. On the other side, the response pathway to signal molecules that are produced by E. coli when it is hungry is deficient in wild-type S. elongatus. To overcome the two defects above, we need to perform genetic manipulation on S. elongatus.

In order to find a suitable strain of S. elongatus, we first learned that S. elongatus PCC7942 and UTEX2973 are constantly used for genetic transformation to study sucrose production and secretion based on previous papers.^{[1, 2]} PCC7942 is easy for gene manipulation but has a relatively slow growth rate. By contraries, UTEX2973 has a relatively high growth rate but is difficult for gene manipulation. We then contacted researcher Xuefeng Lv from Qingdao Institute of Bioenergy and Bioprocess Technology, Chinese Academy of Sciences. We understood and obtained one strain called S. elongatus HL7942, an S. elongatus strain modified from S. elongatus PCC7942. It retains the characteristics of easy gene manipulation and has a higher growth rate than PCC7942 as well. Moreover, it has good genetic stability.

The gene manipulation of S. elongatus HL7942 can follow the natural transformation method. In other words, plasmids can enter S. elongatus cells when we mix bacterial fluid and plasmid solution together. In S. elongatus cells, plasmids cannot be copied. However, those who carry homologous arms of S. elongatus genome can undergo homologous recombination with S. elongatus genome under natural conditions. (Figure 1)

Figure 1. Gene recombination of S. elongatus

Based on the consideration above, we finally chose S. elongatus HL7942 as the strain for design and experiment.

Sucrose permease (CscB) is a sucrose/proton symporter that can utilize proton gradients established across the cell membrane to confer sucrose uptake activity specifically. Considering that most of the heterotrophs tend to acidify the medium, which means S. elongatus is able to secrete sucrose through CscB, especially under hypertonic conditions.^[1] (Figure 2)

Figure 2. Schematic of sucrose permease(CscB) of S. elongatus

In order to simulate the regulatory effect of signal molecules on the expression of CscB, we set the promoter of Cscb gene to lac promoter. In this way, IPTG can qualitatively simulate signal molecules, AHL, to raise the expression of CscB. This will cause the concentration of sucrose in the culture medium to increase. In addition, to select the bacteria that successfully integrate the target fragment, we also designed kanamycin resistance (KanR) gene on the integrated fragment (BBa_K4115045).

Figure 3. Construction of pUC57-NS3-2-Plac_UV5-CscB-lacI-KanR-NS3-2

After the transformation and successful selection of recombinant S. elongatus, we did bacterial colony PCR to prove that the target fragments, especially CscB gene and LacI gene, were integrated into the S. elongatus genome. (Figure 4) Then the target bands were recovered for sequencing. (Figure 5)

Figure 4. Colony PCR of recombinant S. elongatus at different annealing temperature. The area shown in the red box is the target band area. As the annealing temperature rises, non-specific binding reduces. The upper and lower figures are cut from the same gel map for easy annotation and observation.

Figure 5. Sequencing results showed that there was no mutation in CscB gene and LacI gene.

After expanding the culture of recombinant S. elongatus, we set four different cultivation conditions to culture S. elongatus and recombinant S. elongatus. (Table 1) We compared the accumulation of sucrose secreted by a unit number of S. elongatus in a certain period of time by comparing the ratio of sucrose concentration in the supernatant to OD₆₈₅ value of S. elongatus. (Figure 6)

Figure 6. The value of g/(L·OD₆₈₅) of different groups after inducing 24 hours.

We found that under the condition that CscB and NaCl exist together, the effect of sucrose production by S. elongatus is the best. This fact indicates that adding a certain concentration of NaCl to the co-culture medium has a positive effect on our ternary microbial symbiosis system.

We construct intercellular genetic circuit feedback to regulate the nutrient flux in our ternary microbial symbiosis system. In the genetic circuit, starvation promoters in E. coli are sensors for nutrient signals. To ensure a good signal-to-noise ratio for signal input, we made efforts on improving the properties of the starvation promoters.

We evaluate the quality of promoters from two aspects. First, promoter activity is an important evaluation index. Empirically, we believe that promoter activity not lower than the J23101 (a constitutive promoter with moderately strong activity) is necessary for production and genetic circuits. Second, fold-change is another important perspective for promoters, especially for those being used in complicated genetic circuits. Fold-change can be defined as the ratio of the activity in the activated state to the non-activated state. For our starvation promoters, the practical definition is the ratio of promoter activity under low glucose concentration to that under high glucose concentration.

We first selected three starvation promoters and test their response to low glucose concentrations. We construct reporter genes as the construct in Figure 7, and quantified the promoter activity with fluorescence intensity. Among the three promoters, PcstA had the largest fold-change and was comparable to the activity of J23101. However, compared with the widely used inducible promoters like pLac or pBAD, its activity and fold-change are not good enough. So we did several improvement cycles based on the 'Design-Build-Test-Learn' cycle.

Figure 7. sfGFP with LVA is used to indicate the activities of PcstA variants.

PcstA (BBa_K118011) is a well-characterized starvation promoter in iGEM Registry. PcstA uses RpoD as its sigma factor. cAMP receptor protein (CRP) is an activator of PcstA. When undergoes carbon source starvation conditions, the intracellular cAMP concentration will increase and further activate CRP. Then CRP binds to a CRP-binding sequence located upstream of the transcription start site and activates of transcription initiation of PcstA (Figure 8).

Figure 8. CRP-dependent activation of PcstA under starvation conditions.

Furthermore, PcstA is down-regulated by the factor for inversion stimulation (Fis) under nutrient-rich conditions (Figure 9). Under starvation conditions, the Fis abundance in cells will decrease, and the inhibition will be removed.

Figure 9. Fis-dependent inhibition of PcstA under nutrient-rich conditions.

To further optimize the activity and fold-change of PcstA, we made several mutations in its regulation-related sequence. We first mutate the CRP-binding sequence to improve the affinity of the promoter for CRP, so that CRP can activate the promoter more efficiently. The mutants constructed in the first cycle are shown in Figure 4. In principle, if the CRP-binding sequence of PcstA has higher similarity to the consensus sequence, PcstA will have a higher affinity to CRP. The consensus sequence of the CRP-binding site is AAATGTGA-N6-TCTCATTT. The nucleotides with underline are the core of the CRP-binding sequence. G and C with underlines provide most of the binding energy^[4]. Based on these rules, we constructed PcstA-Con and PcstA-TCACA to test what will happen if PcstA has a higher affinity to CRP.

Figure 10. The CRP-binding sequences of PcstA variants in the first cycle.

The reporter gene in Figure 1 is used to test the properties of these PcstA variants. Data is summarized in Figure 11. The promoter activity was significantly improved by the mutations. PcstA-Con has a 15 times higher activity compared with the original PcstA. So it's hard to indicate the data in a column figure. However, it seems that their affinities to CRP are too high. Even under a high glucose concentration of 4 g/L, they are still activated and show no response to starvation. In this cycle, we learnt that high similarity with the consensus sequence is not always a good thing. There might be a suitable range for the affinities of regulators.

Figure 11. Summarized data of the relative activities and fold-changes of PcstA variants. The relative activities are normalized with FI/OD₆₀₀ of J23101 at 4g/L glucose.

Then we decide to mutate another site that has relatively little effect on affinity.

Figure 12. The CRP-binding sequences of PcstA variants in the second cycle.

The same conditions and analysis method were used to test the property of PcstA_Mutant1. The reporter gene constructs are cloned into pUC high-copy number backbone. A significant promoter activity increase was observed on PcstA_Mutant1 and the starvation response function remained (Figure 13A). However, some unwanted phenomena also appeared. First, the leakage expression of PcstA_Mutant1 under high glucose concentration increases, which results in the decrease of fold-change. Second, the high copy of PcstA_Mutant1 in cells has some side effects on growth (Figure 13B). Actually, the growth rate decrease was more obvious in PcstA-Con and PcstA-TCACA transformed strains. So we hypothesize that the high copy of PcstA_Mutant1 causes these side effects. If PcstA_Mutant1 keeps a high copy in cells, it may compete the CRP with the endogenous CRP-dependent promoter and influence the global metabolism.

Figure 13. (A) FI/OD₆₀₀ of PcstA and PcstA_Mutant1 (B) Growth curve of Empty and PcstA_Mutant1 strain in M9 minimal medium containing 4.0 g/L glucose as the sole carbon source.

At the end of the last cycle, we got a function variant, which has higher activity and remains the starvation response. To eliminate the undesirable properties of PcstA_Mutant1, we shifted all constructions to the low-copy number pET backbone and repeat the experiment (Figure 14). The bacteria growth is back to normal in the PcstA_Mutant1 group, which may partially support our hypothesis on the growth. Mutants of PcstA show significantly higher activities as well as larger fold-changes with low-copy number backbone. The data of relative promoter activities and fold changes of PcstA variants in this cycle are summarized in Figure 15 .

Figure 14. FI/OD (600) of PcstA and PcstA_Mutant1. All the constructs are cloned into pET low-copy number backbone.

In conclusion, by mutating the regulation-related sequence and replacing the plasmid backbone, we finally get Pcst_Mutant1 (on pET backbone) as the best construct for the nutrient response.

Figure 15. Summarized data of PcstA variants. High [Glucose]=2 g/L, Low [Glucose]=0.25 g/L. Relative promoter activities are normalized with FI/OD (600) of J23101 at 2 g/L glucose

In A. caulinodans, the nitrogen fixation is tightly controlled by the concentration of intracellular ammonium and this counts for the nitrogenase master regulator nifA^[7]. To achieve controllable regulation of nitrogen fixation rate, endogenous regulators must be removed. In previous research, knocking out of genomic nifA gene and then complementing by nifA controlled by inducible promoters was proved to be a viable strategy^{[5, 6]}. So we first transformed the suicide plasmid into A. caulinodans to construct the ΔnifA strain, and then we transformed the shuttle plasmid into the ΔnifA strain to complement the inducible nifA gene. In the experimental process, we successively carried out two rounds of engineering optimization, finally designed a successful strategy, and completed the verification experiment of the nifA knockout. We learned from research that the knockout of a target gene can be achieved by inserting a DNA sequence into a specific site on the genome using homologous recombination. So we synthesized the 900 bp sequence upstream and downstream of nifA on the genome as the homologous arm (upsteam-homologous-arm, UHA & downsteam-homologous-arm, DHA).

In strategy-one, the inducible nifA fragment (Plac-nifA-ChlR-LacI) containing the ChlR gene, a chloramphenicol resistance marker, was designed to be reversely integrated into the genome (the reintegrated nifA is in the opposite direction to the genomic nifA). In this way, the knockout and inducible complement of nifA could be realized in on step. However, this strategy turned out to failed in PCR validation. Failure to integrate fragments into the genome may be due to excessive length or interference by homologous recombination (nifA itself can also act as a homologous arm). Later we found that it was more likely that the resistance tag failed (seen in strategy-two ).

In strategy-two, only the ChlR gene was designed to be integrated into the genome to knock out the nifA. However, this strategy turned out to failed in PCR validation as well. In subsequent PCR and sequencing analysis, we found that our A. caulinodans ORS571 strain contained ChlR gene on its genome, which which is the information we didn't get through research and inquiry before. Therefore, we concluded that the former two strategies failed due to the incorrect resistance marker.

In strategy-three, the ChlR gene was replaced by bleomycin resistance gene BleoR. And the the inducible nifA fragment was constructed in pBBR1 plasmid with the report fragment (PnifH-sfgfp). Finally, the ΔnifA strain was constructed and validated (seen in Result). Therefore, we designed the suicide plasmid (BBa_K4115041), report shuttle plasmid (BBa_K4115042) and inducible and report plasimid (BBa_K4115043).

Figure 16. Three strategies in controllable regulation of nitrogen fixation rate. (a) Strategy-one; (b) Strategy-two; (c) Strategy-three.

The gel result of the PCR validation preliminarily verified the success of knockout. Further sequencing validation has proved that the construction of ΔnifA strain was completed. More detials and analysis is on Results page.

Figure 17. Colony PCR of ΔnifA colony.

The separate-immobilized fermentation pattern we proposed should be how the three microbial symbiotic system is implemented in practice. To prove this concept, the three microorganisms should survive under some of the same medium condition. We first set up a medium condition in which three microorganisms could separately live.

We had known from previous research about the CoBG11 medium protocols for the coculture of Synechococcus and other heterotrophic microorganisms. Based on this, we proposed our CoBG11-Sucrose medium protocols by adding sucrose as the sole carbon source to simulate the sucrose secretion of S. elongatus. Moreover, ammonium chloride in the medium is used as a simulation of ammonium production by A. caulinodans.

Figure 18. The growth curve of three microorganisms in CoBG11-Sucrose medium.

Through the resulting growth curve, we found that both E. coli and S. elongatus can live and grow in CoBG11-Sucrose medium (Figure 18a&b). However, A. caulinodans quickly reached a stationary phase and remained at a low optical density, which shows that it cannot grow in CoBG11-Sucrose medium (Figure 18.c). After research review, our hypothesis was confirmed. A. caulinodans prefers to use organic acid as its carbon source rather than sugar^[8]. Therefore, when the medium contains only sugar as solo carbon source, A. caulinodans cannot live and grow.

We found that E. coli would excrete significant amounts of acetate when growing aerobically on glucose as the sole carbon source, which SacC-positive E. coli growing on sucrose will be in a similar situation^[9]. In this case, A. caulinodans can take acetate as its solo carbon source. Therefore, we proposed a new CoBG11-Acetate protocol by adding acetate as solo carbon source to simulate the acetate secretion of E. coli. In this experiment, we culture the A. caulinodans both in CoBG11-Sucrose and CoBG11-Acetate to prove that the change of carbon sources can make A. caulinodans successfully live in CoBG11 medium.

Figure 19. The growth curve of A. caulinodans in CoBG11-Sucrose/Acetate medium.

After replacing the carbon source, the growth of A. caulinodans was significantly improved. The result confirms that A. caulinodans can live and grow in CoBG11 medium when acetate is the solo carbon source.

[1] Ducat, D.C., Avelar-Rivas, J.A., Way, J.C., and Silver, P.A. (2012). Rerouting carbon flux to enhance photosynthetic productivity. Appl Environ Microbiol 78, 2660-2668.

[2] Song, K., Tan, X., Liang, Y., and Lu, X. (2016). The potential of Synechococcus elongatus UTEX 2973 for sugar feedstock production. Appl Microbiol Biotechnol 100, 7865-7875.

[3]Lin, P.C., Zhang, F., and Pakrasi, H.B. (2020). Enhanced production of sucrose in the fast-growing cyanobacterium Synechococcus elongatus UTEX 2973. Sci Rep 10, 390.

[4] Soberon-Chavez, G., Alcaraz, L.D., Morales, E., Ponce-Soto, G.Y., and Servin-Gonzalez, L. (2017). The Transcriptional Regulators of the CRP Family Regulate Different Essential Bacterial Functions and Can Be Inherited Vertically and Horizontally. Front Microbiol 8. ARTN 959 10.3389/fmicb.2017.00959.

[5] Ryu, M.H., Zhang, J., Toth, T., Khokhani, D., Geddes, B.A., Mus, F., Garcia-Costas, A., Peters, J.W., Poole, P.S., Ané, J.M., et al. (2020). Control of nitrogen fixation in bacteria that associate with cereals. Nat Microbiol 5, 314-330.

[6] Haskett, T.L., Paramasivan, P., Mendes, M.D., Green, P., Geddes, B.A., Knights, H.E., Jorrin, B., Ryu, M.-H., Brett, P., Voigt, C.A., et al. (2022). Engineered plant control of associative nitrogen fixation. Proceedings of the National Academy of Sciences 119, e2117465119.

[7] Demtröder, L., Pfänder, Y., Schäkermann, S., Bandow, J.E., and Masepohl, B. (2019). nifA is the master regulator of both nitrogenase systems in Rhodobacter capsulatus. Microbiologyopen 8, e921.

[8] 郝保海 (2012). 根瘤菌比较基因组学研究及蛋白质互作网络预测 (华中农业大学).

[9] Bernal, V., Castaño-Cerezo, S., and Cánovas, M. (2016). Acetate metabolism regulation in Escherichia coli: carbon overflow, pathogenicity, and beyond. Applied Microbiology and Biotechnology 100, 8985-9001.