Proof of Concept

The idea

Our project was inspired by the great successes of artificial autotrophy that have been achieved in E. coli (Antonovsky et al., Cell, 2016) or a yeast Pichia pastoris (Gassler et al., Nature Biotechnology, 2020).

To achieve this, we sought to introduce the Calvin Cycle into strains of Streptomyces, a genus of bacteria which produces antibiotics along with other highly valuable molecules and specialized metabolites used in laboratories. To introduce the Calvin Cycle into these bacteria, we sought to transfer the genes encoding RubisCO (Ribulose-1,5-bisphosphate Carboxylase Oxygenase) and PRK (Phosphoribulokinase), as they are the only two missing enzymes that are needed to recreate a variant of the Calvin cycle in heterotrophic organisms (Antonovsky et al., Cell, 2016 ; Gassler et al., Nature Biotechnology, 2020).

However, to meet this challenge, we had to overcome a major problem: the Streptomyces genome has extremely high GC content (72%), which makes the chemical synthesis of genes difficult or impossible. Thus, we had to find a natural source of GC-rich genes encoding RuBisCO and PRK!

Looking for such a natural source of GC rich genes, we made a discovery: a rather poorly described (draft genome) Streptomyces strain, S. bottropensis ATCC 25435, already possesses those two genes! The genes encoding PRK and RuBisCO are located close together and form part of a genomic island surrounded by mobile elements (transposases / recombinases) (Figure 1). This fact suggests that the region might have been acquired via a horizontal transfer from a different species.

Figure 1: Genomic island of S. bottropensis ATCC 25435 encoding notably RuBisCO and PRK

Thus, we thought of using it as a source of genes for our project! The region is about 69% GC rich, thus presenting a close resemblance to genes already present in Streptomyces. Furthermore, often GC rich genes cannot be chemically synthesized, so using these already existing genes seemed to us like the most optimal solution.

Finally, this strain suggests that the implementation of the Calvin cycle in Streptomyces may not be only a dream but really feasible! In order to make this proof of concept, we needed to further characterize S. bottropensis ATCC 25435 strain in absence of carbon source...

CO₂ Fixation

One of the first experiments we carried out was growing different Streptomyces species on a minimal medium containing no carbon source. The results were quite outstanding since we have observed some growth for both S. bottropensis (containing RuBisCO and PRK genes) but also other Streptomyces including S. ambofaciens (Figure 2).

Figure 2: S. bottropensis and S. ambofaciens grown during 4 days on minimal medium devoid of any carbon source at 30°C under standard atmosphere

However, after restriking those colonies on new plates, we did not achieve the same results: the strains did not grow anymore. This result suggests that Streptomyces can initiate a growth in absence of carbon source (maybe thanks to internal carbon stocks) but cannot maintain a prolonged autotrophic growth. We did not get discouraged!

Beyond that, we have also compared the growth of biomass between those same two strains of Streptomyces in standard (0.0415 % CO₂) and CO₂ enriched (3 % CO₂) atmospheres. The final biomass reached by S. bottropensis in sub-minimal liquid medium under 3 % CO₂ enriched atmosphere was twice the final biomass reached in this medium in a standard atmosphere (Figure 3)!

Such a doubling in the biomass under enriched atmosphere was not observed with S. ambofaciens (our control strain) grown in the same conditions. A big step for the project! Perhaps this is a condition for induction of RuBisCO and PRK coding island expression in S. bottropensis.

Figure 3: Biomass reached after 4 days of culture of Streptomyces under standard (0.0415 % CO₂) or CO₂ enriched (3 % CO₂) atmosphere. The media were inoculated with an initial biomass of approximatively 0.04 g.

Since this result is in line with our hypothesis that S. bottropensis contains a functioning carbon-fixating module, we have also decided to perform an assay of the activity of RuBisCO found in this species. We performed a High Performance Liquid Chromatography (HPLC) on cell extracts of S. bottropensis as well as E. coli containing a plasmid with genes encoding the small and large subunits of S. bottropensis RuBisCO codon-optimized for E. coli. We also used an E. coli strain containing a control plasmid.

To measure RuBisCO activity, RuBP (ribulose-1,5-bisphosphate) - a substrate of RuBisCO was added to the samples, to detect the appearance of RuBisCO’s product - PGA (3-phosphoglycerate). To do it, we measured the absorbance of the samples at 220 nm one minute after the addition of RuBP.

Due to a very high baseline in E. coli samples, the results are inconclusive (Figure 4). In S. bottropensis, we detected a small spike attributed to the appearance of PGA in one of the conditions (Figure 5), correlated to the disappearance of RuBP (what is expected for a RuBisCO activity). This result is promising but needs to be confirmed by mass spectrometry analysis (to confirm the identity of the peak attributed to PGA production).

Figure 4: HPLC analysis of RuBisCO assay perform with RuBP (ribulose-1-5-bisphosphate) and cell extract of E. coli harboring a plasmid encoding ccbL (BBa_K4370002) and cbbS (BBa_K4370003) extracts under the control of tetO promoter (BBa_R0040), or control.

Figure 5: HPLC analysis of RuBisCO assay perform with RuBP (ribulose-1-5-bisphosphate) and cell extract of S. bottropensis grown 4 days at 30°C under different conditions.

We have only performed this procedure once, due to time constraints so we suspect a problem with the execution of the experiment.

Finally we characterized the impact on E. coli growth of overexpressing a codon-optimized version of the prk gene of S. bottropensis ATCC 25435 (BBa_K4370004, with an RBS for BBa_K4370005). We observed that the expression of the biobrick is toxic to these cells when arabinose is used as a source of carbon, as previously reported for another PRK. We learnt from this experiment that the genomic island present in S. bottropensis ATCC 25435 encodes a functional PRK enzyme.

Antibiotics production

To find the best chassis for antibiotic production we carried out several experiments to compare different aspects of 7 Streptomyces species, the results of which are summarized in our STREPTObook. All of the strains naturally produce different antibiotics and other secondary metabolites, but we wanted to see which one would be the best candidate to introduce the Calvin Cycle into.

In one of our experiments, we grew all of the seven strains on different media, some of which contained additional CO₂ and then carried out a bioassay, in which we poured an indicator species - Micrococcus luteus around the Streptomyces colony, to observe how Streptomyces inhibits the growth of M. luteus. The results are usually in a form of a halo, the size of which indicates the efficiency of the biostatic effect of the antibiotics produced by Streptomyces (Figure 7).

Figure 6: Antibacterial activity of different Streptomyces strains on sub-minimal medium (devoid of carbon source), on minimal medium containing 0.5% glycerol, or on a rich medium optimized for antibiotic production by S. ambofaciens (MP5 medium containing 36 % glycerol) under regular atmosphere (0.0415 % CO₂) or under 3 % CO₂ ('+CO₂' in the graph). Error bars correspond to the standard error to the mean of two independent experiments.

Figure 7: Antibacterial activity of some Streptomyces on plates with MP5 under regular atmosphere. Scale: the diameter of a plate is 8.5 cm.

From the results (Figure 6, Figure 7), we can see how the antibiotic production differs from strain to strain and medium to medium. For most strains, the antibacterial activity tends to decrease in the presence of 3% CO₂. This decrease remains modest in most cases, so this provides proof of concept that antibiotic production in a CO₂-enriched atmosphere is possible. Overall, from the results we can see that S. rimosus and S. venezuelae are generally more efficient in antibiotic production than other strains.

These results provide a first proof of concept that although lower, antibiotic production is possible in sub-minimal medium (devoid of carbon source) and in CO₂-enriched atmosphere. They also lead us to elaborate strategies to improve antibiotic production under these conditions (see below the sgRNA directed against lsr2A gene). Finally these results also allow us to identify strains that would be better chassis for our project (S. rimosus and S. venezuelae), but which will need to be genetically modified to introduce genes of interest (encoding the RuBisCO and PRK) as well as sgRNA able to decrease the expression of target genes (see below).

Genetic engineering of Streptomyces

For this reason, we performed conjugation assays to test the ability of our strains to be modified following a standard protocol. We performed a conjugation from E. coli to Streptomyces of a test plasmid (pOSV805, Aubry et al., Appl Environ Microbiol, 2019) containing a gene for antibiotic resistance (Figure 8), using a protocol that was previously optimized for S. ambofaciens. All of the strains, except for S. rimosus, had successfully obtained the plasmid. These results indicate that all strains except S. rimosus can be easily genetically modified with an integrative plasmid and therefore are suitable for use as chassis. Despite promising results regarding antibiotic production in low carbon conditions (Figure 6), S.rimosus is not an optimal chassis regarding the conjugation assay. This means that this strain should either be eliminated from our list of potential candidates for the CO2CURE project, or that further experiments should be conducted to optimize the conjugation protocol for this strain. Fortunately, we were able to identify a publication in which an optimized protocol for this species had been published (Song et al., J Zhejiang Univ Sci B. , 2019)). The S. rimosus chassis therefore remains potentially interesting.

Figure 8: Example of conjugation of pOSV805 integrative plasmid into S. venezuelae. The colonies correspond to exconjugants that have acquired the resistance marker (hygromycin resistance).

Gene silencing

Furthermore, to obtain antibiotic production in the sub-minimal medium (devoid of carbon source, Figure 6), we had to repress the expression of the lsr2a gene which is an xenogeneic silencer. This silencer represses the gene clusters such as SMBGCs (specialized metabolite biosynthetic gene clusters) and stops biosynthesis of different metabolites in Streptomyces (Deng et al., RSC Adv., 2017 ; Gehrke et al., eLife, 2019; Zhang et al., mBio, 2021). We have engineered genetic tools via the Crispr-dCas9 system to switch off lsr2a and force antibiotic production in a low carbon environment.

We took advantage of the same methodology (CRISPR-dCas9) to switch off the glycerophosphate mutase genes to force the appearance of the Calvin cycle in their primary metabolism as learned from previous studies performed in E. coli (Antonovsky et al., Cell, 2016).

We have successfully transferred by conjugation the CRISPR-dCas9 plasmids containing guide RNA sequences targeted at both lsr2a and gpm into S. ambofaciens and S. venezuelae species. However, due to time constraints we were able to carry out any experiments only on sgRNA gpm (BBa_K4370012) exconjugants. We noticed that clones containing the BBa_K4370012 biobrick show a slight growth defect, which is indicative of an effect of the presence of the sgRNA. This observation is consistent with the fact that the a decrease in the expression of the phosphoglycerate mutase (encoded by gpm) disrupts carbon flow in the glycolytic/gluconeogenic backbone as previously reported in E. coli (Antonovsky et al., Cell, 2016). However, these exconjugants are still able to produce antibiotics in poor conditions. This is encouraging in the context of an application of this biobrick to facilitate the implementation of the Calvin cycle in Streptomyces for antibiotic production after CO₂ fixation.

Future of the project