Towards a synthetic yeast endosymbiont with a minimal genome

Abstract

Based on the endosymbiotic theory, one of the key events that occurred during mitochondrial evolution was an extensive loss of non-essential genes from the protomitochondrial endosymbiont genome and transfer of some of the essential endosymbiont genes to the host nucleus. We have developed an approach to recapitulate various aspects of endosymbiont genome minimization using a synthetic system consisting of E. coli endosymbionts within host yeast cells. As a first step, we identified a number of E. coli auxotrophs of central metabolites that can form viable endosymbionts within yeast cells. These studies provide a platform to identify non-essential biosynthetic pathways that can be deleted in the E. coli endosymbionts to investigate the evolutionary adaptations in the host and endosymbiont during the evolution of mitochondria.

Graphical Abstract

The endosymbiotic theory¹ of mitochondria evolution suggests that mitochondria evolved from prokaryotic lineages that entered an archaeal host and were established as endosymbionts.^2–3 Several phylogenetic and proteomic studies suggest that all mitochondrial genomes are descended from a common protomitochondrial ancestor.^4–8 One of the key events that occurred during mitochondrial evolution is an extensive loss of non-essential genes from the protomitochondrial endosymbiont genome and transfer of some of the essential endosymbiont genes to the host nucleus.⁹ This resulted in the relatively low gene content of the present-day mitochondrial genome. Nature has used similar strategies to shrink genomes in various intracellular pathogenic organisms, and ground-up approaches to synthetically obtain minimal genomes have also been performed on intracellular pathogens like Mycoplasma.^10–13

Recently, we developed a synthetic strategy to establish E. coli endosymbionts within host yeast cells to mimic mitochondrial evolution starting from bioenergetically competent and genetically tractable endosymbionts.¹⁴ Briefly, our strategy (Figure 1) involved engineering E. coli strains such that they (i) are auxotrophic for a key central metabolite (present in the host cytosol) for which there are endogenous transporters, (ii) express a functional ADP/ATP translocase to provide ATP to the respiration deficient S. cerevisiae mutant and (iii) express SNARE-like proteins to avoid lysosomal degradation as previously described.^14–18 A respiration deficient mutant of S. cerevisiae, S. cerevisiae cox2–60 (NB97), was used which is unable to synthesize ATP under non-fermentable growth conditions.^19–21 Under selection conditions containing non-fermentable carbon sources, the S. cerevisiae cox2–60 obtains ATP from the engineered E. coli endosymbiont and the S. cerevisiae cox2–60 cytosol in return provides thiamin to the E. coli endosymbiont that is a thiamin auxotroph. The logical next step would be to expand this system towards a minimal endosymbiont genome, similar to mitochondria evolution. We envisioned that the yeast cytosol would provide an ideal “rich medium” for the growth of E. coli endosymbionts thereby allowing us to delete numerous non-essential genes and pathways under these conditions.^22–24

Figure 1: Strategy to test permissive E. coli auxotrophs that could establish endosymbiosis with S. cerevisiaecox2–60.

X- key central metabolite provided by the yeast cytosol, M-mitochondria, N-nucleus, ER-endoplasmic reticulum, G-Golgi apparatus, V-vacuole.

Since amino acid biosynthesis is a major component of bacterial metabolism, we began by testing E. coli auxotrophs for the twenty amino acids.²⁵ Similar to the selection platform described above, the S. cerevisiae cox2–60 should obtain ATP from the engineered E. coli endosymbiont that is auxotrophic for a defined amino acid(s), and in return, the S. cerevisiae cox2–60 cytosol was expected to provide the corresponding amino acid(s) to the E. coli endosymbiont. Since E. coli encodes amino acid transporters, the E. coli endosymbiont should be able to acquire amino acids from the yeast cytosol. To this end, E. coli strains that were individual amino acid auxotroph were transformed with a plasmid - pAM136,¹⁴ containing ADP/ATP translocase from the intracellular bacterium Protochlamydia amoebophila strain UWE25, and SNAREs - Chlamydia trachomatis incA, Chlamydia caviae incA, Chlamydia trachomatis CT_813 under a P_BAD promoter. These strains were then fused to respiration deficient S. cerevisiae cox2–60 (NB97). The S. cerevisiae cox2–60-E. coli chimeras were selected on partial selection medium I (3% glycerol and 0.1% glucose as carbon source) containing predominantly non-fermentable carbon sources. The permissive E. coli amino acid auxotrophs that were able to establish endosymbiotic relationship are listed in Table 1.

Table 1:

List of E. coli auxotrophs that were tested as for their ability to establish endosymbiotic relationship with S. cerevisiae cox2–60

Strain #	Gene deleted/Strain	Auxotroph	Growth on selection medium I & II
1	ilvD	isoleucine, leucine, valine	++
2	trpC	tryptophan	+
3	serA	Serine	++
4	metA	methionine	++
5	PA340 strain	glutamic acid	++
6	cysE	cysteine	++
7	glnA	glutamine	++
8	argA	arginine	++
9	proA	proline	++
10	lysA	lysine	++
11	thrC	threonine	++
12	hisB	histidine	+
13	tyrA	tyrosine	−
14	pheA	phenylalanine	−
15	DL39 strain	phenylalanine, tyrosine, aspartic acid, leucine, isoleucine, valine	−
16	KA12 strain	Aromatic amino acids	−
17	thiC	thiamin	++
18	nadA	NAD	++
19	thiC, nadA	thiamin and NAD	++
20	thiC, nadA, metA	thiamin, NAD, methionine	++
21	thiC, nadA, metA, ilvD	thiamin, NAD, methionine, leucine, isoleucine, valine	++
22	thiC, nadA, metA, ilvD, argA	thiamin, NAD, methionine, leucine, isoleucine, valine, arginine	++
23	thiC, nadA, metA, ilvD, argA, proA	thiamin, NAD, methionine, leucine, isoleucine, valine, arginine, proline	++
24	thiC, nadA, metA, ilvD, argA, proA, glnA	thiamin, NAD, methionine, leucine, isoleucine, valine, arginine, proline, glutamine	+
25	thiC, ΔilvLXGMEDAYC	thiamin, leucine, isoleucine, valine	++

As can be seen from these experiments, most of the E. coli single amino acid auxotrophs tested are permissive endosymbionts. All of the chimeric strains generated from the permissive E. coli auxotrophs and S. cerevisiae cox2–60 were able to grow on selection medium II (containing only non-fermentable carbon sources, 3% glycerol and no glucose) for multiple rounds of re-plating and re-growth (> 40 generations) unlike the host S. cerevisiae cox2–60. Growth rates for all permissive auxotrophs listed in Table 1 and Figure S19 (except for histidine and tryptophan auxotrophs) were similar to those of S. cerevisiae cox2–60 – E. coli DH10B ΔnadA::gfp-kan^R- pAM136 (Figure S19).¹⁴ Significantly slower growth rate (>12 h per doubling on selection medium I) was observed for S. cerevisiae cox2–60 – E. coli DH10B ΔtrpC::kan^R- pAM136 and S. cerevisiae cox2–60 – E. coli DH10B ΔhisB::kan^R- pAM136. Further, we isolated total genomic DNA from the chimeric strains and PCR amplified characteristic marker gene MATa from S. cerevisiae cox2–60 and kan^R gene corresponding to the E. coli auxotrophs to confirm the presence of E. coli endosymbionts within S. cerevisiae cox2–60 (see SI for data).

The inability to form endosymbionts with the Phe and Tyr auxotrophs could be due to lower intracellular concentration of phenylalanine and tyrosine in the yeast cytosol^26–27. Unfortunately, our efforts either to add tyrosine to the growth medium or over-express aromatic amino acid uptake systems in E. coli (e.g., aroP) were not able to rescue this phenotype. Likewise, at present it is unclear why slower growth rates were observed for E. coli tryptophan and histidine auxotrophs (reported intracellular concentrations in the yeast cytosol: tryptophan ~ 450 μM and histidine ~ 9 mM^26–27).

Because we ultimately want to minimize the E. coli genome, we asked whether we could delete the entire operon for one of the permissive auxotrophs. To test this notion, we used a deletion mutant of the isoleucine-valine biosynthetic pathway. We transformed this E. coli strain (strain 25, Table 1) with pAM136, fused it to S. cerevisiae cox2–60, and selected for S. cerevisiae cox2–60 – E. coli ΔthiC::kan^R-gfp ΔilvLXGMEDAYC::tet^R chimeras on selection medium I (Figure S18). We were able to propagate the chimeras on selection medium I and II for multiple rounds of re-growth, and the chimera grew at a comparable rate to that of S. cerevisiae cox2–60 – E. coli DH10B ΔnadA::gfp-kan^R- pAM136 (~ 6 h per doubling on selection medium I).¹⁴

Next, we wanted to investigate whether we could generate E. coli strain auxotrophic to multiple metabolites and test the ability of such auxotrophic strains to generate yeast-endosymbionts. We had previously tested thiamin and NAD auxotrophs of E. coli and both of these auxotrophs along with their combination were permissive E. coli endosymbionts.¹⁴ We therefore started with a thiamin/NAD double auxotroph of E. coli generated by deleting the thiC and nadA genes. We further deleted the metA gene from this strain to generate E. coli strain 20 (Table 1) that is auxotrophic for thiamin, NAD and methionine.²⁸ As before, this strain was transformed with pAM136 and fused to S. cerevisiae cox2–60. We were able to generate chimeras that propagated on selection medium II (3% glycerol and no glucose) for more than 40 generations of growth, and the growth rate was similar to that of S. cerevisiae cox2–60 – E. coli DH10B ΔnadA::gfp-kan^R- pAM136 (Figure S19).¹⁴

Next, we sequentially generated a series of E. coli strains that were auxotrophic in an increasing number of metabolites up to an E. coli strain auxotrophic to 8 different metabolites (thiamin, NAD, methionine, isoleucine, leucine, valine, arginine and proline, strain 23. All of these auxotrophic strains (strains 20–23 transformed with pAM136) were permissive yeast endosymbionts. The chimeras were able to sustain growth on both Selection Medium I and Selection Medium II, thereby demonstrating a respiration competent phenotype unlike the host strain S. cerevisiae cox2–60 (e.g., Figure 2A and 2B respectively, for chimeras generated from strain 23 of E. coli). Moreover, these chimeric strains showed robust growth on selection medium I (Figure S19).¹⁴ We also isolated the total genomic DNA from the chimeric strains (generated from strains 20–23) and PCR amplified the characteristic marker gene MATa from S. cerevisiae cox2–60 and gfp/tet^R gene from the E. coli endosymbiont to confirm the presence of the endosymbionts within S. cerevisiae cox2–60 (Figure 2C–F, Figure S13–16). Furthermore, as a representative endosymbiont, the S. cerevisiae cox2–60 - E. coli ΔthiC:gfp ΔnadA ΔmetA ΔilvD, ΔargA, ΔproA – pAM136 chimera was characterized by fluorescence imaging to confirm the presence of the E. coli endosymbionts within the yeast. As can be seen from Figure 2G, we detect the presence of GFP expressing endosymbionts within the host S. cerevisiae cox2–60 cells.

Figure 2: Characterization of Saccharomyces cerevisiae cox2–60 - E. coli DH10B ΔthiC:gfp ΔnadA ΔmetA ΔilvD, ΔargA, ΔproA – pAM136 chimeras.

(A) and (B), Growth observed for chimeras on Selection Medium I and Selection Medium II respectively, thereby demonstrating a respiration competent phenotype. (C) and (D), Detection of NB97-specific MATa mating type by PCR analysis of total DNA samples isolated from chimeras shown in Figure 2A and 2B respectively. Std indicates DNA molecular weight standards (E) and (F) Detection of gfp gene by PCR analysis of DNAs isolated from chimeras shown in Figure 2A and 2B respectively. (G) Detection of GFP expressing E. coli endosymbionts within Saccharomyces cerevisiae cox2–60. No GFP signals are detected in control Saccharomyces cerevisiae cox2–60 cells. Scale bars represent10 μm.

When we started deleting additional genes (e.g., glnA, strain 24) from strain 23 - E. coli ΔthiC:gfp ΔnadA ΔmetA ΔilvD, ΔargA, ΔproA, we observed significantly slower growing chimeras. As a first step, we genome sequenced strain 24 to investigate if there were any genomic rearrangements that might have occurred in strain 24 during our recombineering efforts²⁸ that might have resulted in slower growth of chimeras. We did not detect any such rearrangements when compared to E. coli ΔnadA:gfp that we had previously generated. At this point, we are pursuing E. coli genomic library complementation²⁹ and mutation/selection based evolutionary approaches to attempt to correct the growth defects in the chimeras. Similar approaches are being tested to correct the growth defects in case of chimeras generated from E. coli strains auxotrophic for tyrosine, phenylalanine, tryptophan and histidine.

In conclusion, we were able to identify a number of E. coli auxotrophs corresponding to central metabolites (i.e., amino acids and cofactors) as potential endosymbionts within yeast cells. Further, we generated an E. coli strain auxotrophic to 8 different metabolites, (for which the biosynthetic pathways involve about 40 gene products ~ 46 kb³⁰), and demonstrated the generation of S. cerevisiae cox2–60-E. coli chimeras using this strain. Considering that more than 30 percent of genes in E. coli are functionally assigned to metabolism³⁰, these studies are expected to set up a platform towards the deletion of a significant fraction of non-essential genes in the E. coli endosymbionts in the context of the yeast cytosol. Our eventual goal would be to obtain a minimal E. coli genome within yeast cells. Such studies are expected to provide insights into the evolutionary adaptations that may have been necessary for mitochondrial evolution starting from prokaryotic endosymbionts.