Tailoring the evolution of BL21(DE3) uncovers a key role for RNA stability in gene expression toxicity | Communications Biology – Nature.com

Restricting the evolutionary space of BL21(DE3) for the production of toxic proteins

In DE3 strains, the strong lacUV5 promoter drives expression of T7 RNAP, and it was previously shown that homologous recombination between lacUV5 and the weaker wild-type lac promoter is a dominating event that within hours of introducing a toxic gene leads to tolerance in BL21(DE3)15. To prevent this, we first genetically restricted BL21(DE3), creating a lacIlacZ variant by deleting exactly the part of the native lac locus that shared homology with the DE3 locus (Fig.1). In addition, this should prevent the expression of lacY, encoding the lactose permease, allowing a more uniform, concentration-dependent entry of IPTG into all cells in bacterial populations16.

a Illustration showing the experimental set-up used to restrict the evolutionary space for BL21(DE3) to overcome protein production toxicity on three levels: (upper panel) genetic restriction was accomplished by deleting a part of the BL21(DE3) genome that frequently recombines and lowers the T7 RNAP expression, (middle panel) protein production was coupled to both fluorescence and antibiotic resistance to prevent the formation of non-producing mutants and (lower panel) a spatiotemporally structured environment of ageing bacterial colonies was designed in order to identify strong phenotypic mutants. b Schematic of the workflow to produce layered agar plates that allow diffusion of the antibiotic ampicillin (AMP)and the inducer IPTG to allow sufficient time for colony formation on the surface of the top agar layer before toxic protein production is induced.

Next, similar to a previous study12, we chose the E. coli membrane protein insertase YidC coupled to GFP to serve as a model protein to investigate stress caused by membrane protein overproduction as it formerly was shown to have a strong negative fitness effect on E. coli17,18. The GFP fusion comes in handy for phenotypic restriction because it allows visual screening for mutants that still produce the fusion protein. We further restricted the evolutionary solution space by introducing a hairpin structure in the expression plasmid that couples YidC-GFP to the expression of a -lactamase gene19, providing resistance to ampicillin (Fig.1). This way, the formation of non-producing populations should be minimised in the presence of the antibiotic.

Finally, in contrast to the previous studies11,13,20, we aimed at introducing the protein production stress after bacterial colonies had established on plates. This spatiotemporally different approach should allow the formation of a large number of bacteria in a state of dormancy and in a structured environment, which previously was shown to constitute a unique evolutionary environment21. Furthermore, we speculated that mutants would be easy to identify as fluorescent secondary colonies, so-called papillae, outgrowing from the initially established colonies. To this end, two-layered agar plates were poured, allowing slow IPTG and ampicillin diffusion from the bottom to the top layer to grant sufficient time for colony formation before YidC-GFP production was induced (Fig.1).

During incubation for 1 week at 37C, we observed several fluorescent papillae (Fig.1) that were restreaked to confirm the fluorescent and ampicillin-resistant phenotype. Based on the fluorescence phenotype, the most promising mutants were isolated and cured of the YidC-GFP plasmid using a mild and simple CRISPR-based approach to plasmid curing22. These strains were subsequently retransformed with the original YidC-GFP plasmid to ensure that mutations leading to tolerance were localised on the genome and not on the expression plasmid. A single clone (Evo21(DE3)) was chosen for further characterisation.

To benchmark Evo21(DE3), we compared its ability to produce the YidC-GFP-fusion protein to the non-evolved BL21(DE3) wild-type strain as well as the derivative Mt56(DE3) previously described to be optimised for YidC-GFP overproduction12. The plasmid pLysS was utilised to limit basal T7 RNAP expression23 in BL21(DE3) and Evo21(DE3). Co-expression of pLysS in Mt56(DE3) is not relevant due to the greatly reduced polymerase activity in this strain. In liquid culture, based on GFP fluorescence, Evo21(DE3) expressed significantly (P0.0001) higher amounts of YidC-GFP than BL21(DE3) and Mt56(DE3). Eight hours post induction, yields were 3.6-fold higher in cultures of Evo21(DE3) than in cultures of BL21(DE3) and 2.1-fold higher than in Mt56(DE3) (Fig.2a). Parallel monitoring of the optical density of the cultures showed no major growth impairment for the strains (Supplementary Fig.1).

a Production of a toxic YidC-GFP-fusion protein in Evo21(DE3) compared to the non-evolved BL21(DE3) wild-type strain, the BL21(DE3)lacI/Z ancestor strain, and a previously evolved BL21(DE3) derivative Mt56(DE3). On the illustrated expression vector (pYidC), YidC-GFP production is translationally coupled to ampicillin resistance. b Fluorescence fold change between Evo21(DE3) and BL21(DE3), producing a library of 24 proteins of the E. coli inner membrane proteome C-terminally fused to GFP. c gfp expression levels in Evo21(DE3) and control strains with and without co-expression of the helper plasmid pLysS. d Western blot showing the expression of a camelid-derived single-chain antibody (nanobody) in Evo21(DE3) and control strains. Samples were normalised to cell density before loading. e Production of YidC under the control of a PrhaBAD promoter allowing titration of expression l-rhamnose. All fluorescence values displayed are normalised to OD at 600nm. Error bars indicated represent the average squared deviation from the mean (SD) of three biologically independent samples (n=3).

To assess whether the Evo21(DE3) phenotype was gene-specific, we next investigated the expression of a set of 24 different GFP-fusion proteins selected from an expression vector library of the E. coli inner membrane proteome24. These membrane proteins were selected to cover a wide range of functions, toxicity (previously reported as a change in OD600 upon IPTG addition24), and the number of predicted transmembrane domains (Supplementary Table1). Comparing the fold change of protein production in Evo21(DE3) and BL21(DE3), titre was improved for 19 of the 24 proteinswith a significant fold change of more than 1.5-fold for 14 of them (Fig.2b). The highest improvement was 6.1-fold (P0.001), observed for the protein YihG.

As a first test that the underlying mechanism allowing Evo21(DE3) to produce more toxic protein was not related to a general decrease in the activity of the T7 expression systemas previously observed for the BL21(DE3) derivatives C41/43(DE3), C44/45(DE3) and Mt56(DE3)we Sanger sequenced the T7 RNAP gene, which confirmed an absence of mutations. Next, we compared the expression of two non-toxic soluble proteins, GFP and a camelid-derived nanobody, and both were produced at higher levels in Evo21(DE3) than in BL21(DE3) or Mt56(DE3)both in the presence and absence of pLysS (Fig.2c, d). Similarly, Evo21(DE3) outperformed other strains in the production of seven out of eight different plant-derived cytochrome P450 enzymesa class of enzymes of significant biotechnological interest25 (Supplementary Fig.2). This indicates that the causative mutation in Evo21(DE3) is different from previously isolated BL21(DE3) derivatives and probably does not cause a general decrease in T7 RNAP activity.

Even though the screening for improved protein productivity was performed with the highly efficient T7 system, the ideal mutant strain would be capable of producing more protein independently from the promoter system. To explore if this was the case for Evo21(DE3), we replaced the T7 promoter with the l-rhamnose inducible rhaBAD promoter in the yidC-GFP expression vector, transformed it into Evo21(DE3), and expressed the construct by inducing with different rhamnose concentrations in liquid culture. With concentrations of 5 and 20mM l-rhamnose, Evo21(DE3) produced significantly (P0.0001) more protein than BL21(DE3) and Mt56(DE3) (Fig.2e).

In summary, this initial characterisation shows that the evolved strain can produce a higher titre of a range of different proteins using a T7-system-independent mechanism.

The phenotype of Evo21(DE3) prompted us to sequence the strain using Illumina whole-genome sequencing. Two point mutationsone in the argE and one in the fecB locus (Fig.3a)and an insertion of a mobile IS1 element into the rne gene were identified. Upon reintroduction of the argE or fecB point mutations into BL21(DE3) by oligonucleotide-based recombineering, the YidC-GFP overexpression phenotype was not obtained (data not shown), whereas, when reintroducing the truncation of the rne locus into BL21(DE3) and the Evo21(DE3) parental BL21(DE3) lacIlacZ strain, the YidC-GFP overexpression tolerance phenotype was nearly identical to Evo21(DE3) (Fig.3b). This makes it highly likely that the rne IS1 insertion is the main causative mutation in Evo21(DE3).

a Illustration of mutations in the Evo21(DE3) genome compared to the ancestor BL21(DE3). Whole-genome sequencing of mutant strain Evo21(DE3) revealed two point mutations (grey) and a truncation of the rne gene caused by the transposition of a mobile element IS1 into the locus. The deletion of the genomically shared homology sequence in the BL21(DE3) ancestor strain with the DE3 area (335,401337,123) is annotated. b Production levels of the toxic YidC-GFP-fusion protein in Evo21(DE3) compared to BL21(DE3), as well as BL21(DE3) and the non-evolved ancestor strain BL21(DE3)lacIlacZ after reintroducing the rne truncation by recombineering. Error bars indicated represent the average squared deviation from the mean (SD) of three biologically independent samples (n=3). c Illustration of the E. coli RNA degradosome. N- and C-terminal domain of the membrane-bound essential endonuclease RNase E (blue) and the localisation of associated enzymes PNPase, Rhlb and enolase along the C-terminal non-catalytic scaffolding region are displayed. Mutations of the rne gene in Evo21(DE3), BL21Star(DE3) and rne* gene harboured on pLysS-Max are indicated.

The identified IS1 insertion causes a truncation of the encoded 1061-residue E. coli endoribonuclease RNase E after amino acid 702 and, therefore, a polypeptide lacking the last 359 residues of its C-terminus in Evo21(DE3) (Fig.3c). RNase E is an essential membrane-associated enzyme involved in the maturation of both ribosomal RNA and tRNA, as well as total mRNA decay, and mediates the assembly of a multi-enzyme complex referred to as the RNA degradosome (Fig.3c). It has previously been shown that only the N-terminal half (1529) of RNase E, accommodating the active catalytic domain, is essential for cell growth, and the C-terminal non-catalytic region is mostly disordered and known to function as a scaffold mediating the association of the enzymes PNPase, Rhlb and enolase26,27,28.

Interestingly, a similar truncation of the rne locus, rne131, resulting in an RNase E polypeptide lacking its non-catalytic region (amino acid residues 1584, Fig.3c) was isolated in a screen for suppressors of a temperature-sensitive allele of the mukB gene26. A later study showed that in strains such as BL21(DE3), introducing the rne131 truncation caused a bulk stabilisation of mRNA degradation, including mRNA produced by T7 RNAP29. The rne131 truncation was engineered into the commercially available BL21Star(DE3) with the rationale that stabilising bulk mRNA would result in increased protein production. However, following the same rationale, the commercial strain also comes with a note suggesting that it might be unsuitable for overexpression of toxic genes.

We compared the expression of six different genes that we previously found expressed better in Evo21(DE3) than in BL21(DE3) with expression in BL21Star(DE3) and found expression levels to be highly similar between BL21Star(DE3) and Evo21(DE3) (Fig.4a). This provides an independent confirmation that the phenotype of Evo21(DE3) is caused by the truncation of rne.

a Heterologous production of a variety of non-toxic and toxic GFP-fusion proteins in Evo21(DE3) and BL21Star(DE3), both harbouring different truncations of the rne gene, compared to BL21(DE3). b Expression of the same genes in Evo21(DE3) is compared to expression in BL21(DE3) when co-expressing either the auxiliary plasmid pLysS or pLysS-Max. c Schematic illustration of the plasmid pRNE-GFP designed to report on RNase activity. Promoters controlling rne expression are indicated (P14)49,50. The expression level of the rne gene can be monitored in vivo via GFP fluorescence signal. d Exploration of the pRNE-GFP reporter plasmid in BL21(DE3) derivatives. Different levels of RNase activity can be monitored in the strains. e Titration of yidC expression in Evo21(DE3), BL21Star(DE3) and BL21(DE3) via increasing levels of l-rhamnose and its effect on the rne regulon expression reporting RNase activation in the cell during toxic membrane protein production. Upper half: Fluorescence corresponds to YidC-GFP production levels. Lower half: Cells harbour both the rne reporter plasmid (pRNE-GFP) and pPrhaBAD-YidC controlling yidC expression (no GFP fusion). Fluorescence levels correspond to GFP produced under the control of the rne regulon. f Effect of the pLyS-Max auxiliary plasmid on RNase activity. Plasmids pLysS and pLysS-Max are co-expressed along with pRNE-GFP in BL21(DE3) cells. To repress leaky expression of the rhamnose promoter controlling rne* expression on pLysS-Max, 0.4% glucose was added where indicated. Error bars indicated represent the average squared deviation from the mean (SD) of three biologically independent samples (n=3).

The way Evo21(DE3) was isolated from papillae outgrowing colonies on week-old agar plates, and because dominant rne mutants previously have been observed30, made us speculate that different rne variants could be studied by simple co-expression from a plasmid in the presence of wild-type rne on the genome. To test this idea and to compare different variants of the rne gene at different expression levels, we cloned rne variants in front of the rhaBAD promoter on the pLysS plasmid backbone: full-length rne, the Evo21(DE3) and BL21Star(DE3) truncated versions and a version with a further truncation in the membrane-binding domain of RNase E. However, none of these constructs showed any positive effect on YidC-GFP expression (Supplementary Fig.3).

Serendipitously, we isolated a spontaneously occurred rne mutant, called rne*, and included it in our analysis. We found that rne* provided on the pLysS plasmid (hereafter called pLysS-Max) could increase YidC-GFP production even further than Evo21(DE3) (Fig.4b). The rne* mutation converts the essential31 aspartate residue in position 346 to an asparagine in the so-called DNase I subdomain of RNase E involved in chelating an essential Mg2+ ion. The aspartate residue is believed to act as a general base to activate the attacking water essential for the catalytic activity of the enzyme32. The replacement of Asp-346 with the polar amino acid Asn was previously shown to decrease RNA cleavage by about 25-fold32. The effect of expressing rne* was not gene-specific as the effect was preserved for three out of four other tested genes (Fig.4b). This provides an alternative demonstration that manipulating with rne severely affects recombinant protein production and provides a simple tool, in the form of an auxiliary pLysS-Max plasmid that can be transformed into other strains, to improve protein production titre.

The positive effect of rne truncations such as rne131 on protein production was previously assumed to be due to the stabilisation of the recombinant mRNA29. However, the observation that the rne truncation in Evo21(DE3) leads to tolerance of toxic gene expression suggests a broader role involving balancing of RNA levels more globally.

Autoregulation allows RNase E to continuously adjust its synthesis to that of its substrates by controlling the degradation rate of its own mRNA33,34. This could work as a biosensor for RNase E activity by fusing the promoter and 5end of rne with a genetic reporter, as previously demonstrated with lacZ34. To explore RNase E activity in our evolved strains for recombinant protein production, we constructed a similar RNase E biosensor (pRNE-GFP) using GFP as a reporter (Fig.4c).

We transformed the pRNE-GFP reporter into BL21, BL21(DE3), BL21Star(DE3), Evo21(DE3) and Mt56(DE3) and monitored fluorescence in a microplate reader under conditions similar to the typical protein production scenario. Surprisingly, under these conditions, fluorescence was 14-fold reduced in Evo21(DE3) and sevenfold reduced in BL21Star(DE3) compared to the ancestral BL21(DE3 (Fig.4d). Given that Evo21(DE3) behaves almost identically to BL21Star(DE3) and that the rne131 truncation has been shown to cause a bulk stabilisation of mRNA degradation29, this suggests that low fluorescence from our reporter correlates with increased bulk mRNA stabilisation. Interestingly, fluorescence from the reporter was reduced approximately twofold in BL21 compared to BL21(DE3), suggesting an effect of the DE3 locus itself on RNase E activity in the cell.

Next, we wanted to explore if the expression of YidC-GFP affected RNase E activity in the different strains. To this end, because YidC-GFP cannot be expressed from the T7 promoter in BL21 (no T7 RNA polymerase) or BL21(DE3) (no growth), we expressed it from the rhaBAD promoter construct (Fig.2e) using different concentrations of rhamnose. This showed that YidC-GFP levels could be titrated with rhamnose and confirmed higher expression in BL21Star(DE3) and Evo21(DE3) than in the other strains at high rhamnose concentrations (Fig.4e, upper half).

To monitor fluorescence from the RNase E GFP reporter, we then deleted GFP from the YidC construct and attempted to co-transform it with pRNE-GFP into different strain backgrounds (Fig.4e, lower half). However, we were unable to recover and grow transformants in BL21(DE3) and BL21Star(DE3), suggesting a lethal imbalance in RNA levels caused by the presence of these two plasmids. We were able to recover and grow double transformants in BL21, Evo21(DE3), and Mt56(DE3) and monitor fluorescence as an indication of RNase E activity. Fluorescence increased to high levels upon increasing the concentration of rhamnose in BL21 and Mt56(DE3), but fluorescence levels were at least 2.5-fold lower (at 1mM rhamnose) in Evo21(DE3) and hardly increased upon further rhamnose addition. This suggests that RNase E activity towards bulk mRNA is increased when yidC expression is increased but that the activity is lower in Evo21(DE3) than in the other strains (Fig.4e).

Finally, we explored the effect of the pLysS-Max auxiliary plasmid (harbouring the mutant rne*) on RNase E activity by co-transforming it along with pRNE-GFP into BL21(DE3). As controls, we included a strain co-transformed with pLysS and pRNE-GFP and BL21(DE3) transformed only with pRNE-GFP. Because we previously observed effects on RNase E activity due to the presence of the DE3 locus, we repressed leaky expression of T7 RNAP from the lacUV5 promoter by adding glucose to the medium and titrated rne* levels with rhamnose. In the absence of glucose and the presence of pLysS, we observed an increase in fluorescence which was repressed by expression of rne* (Fig.4f). This shows that the pLysS-Max plasmid can be used to regulate RNase E activity in the cell.

Continue reading here:

Tailoring the evolution of BL21(DE3) uncovers a key role for RNA stability in gene expression toxicity | Communications Biology - Nature.com