Prolonging genetic circuit stability through adaptive evolution of overlapping genes

Abstract The development of synthetic biological circuits that maintain functionality over application-relevant time scales remains a significant challenge. Here, we employed synthetic overlapping sequences in which one gene is encoded or ‘entangled’ entirely within an alternative reading frame of another gene. In this design, the toxin-encoding relE was entangled within ilvA, which encodes threonine deaminase, an enzyme essential for isoleucine biosynthesis. A functional entanglement construct was obtained upon modification of the ribosome-binding site of the internal relE gene. Using this optimized design, we found that the selection pressure to maintain functional IlvA stabilized the production of burdensome RelE for >130 generations, which compares favorably with the most stable kill-switch circuits developed to date. This stabilizing effect was achieved through a complete alteration of the allowable landscape of mutations such that mutations inactivating the entangled genes were disfavored. Instead, the majority of lineages accumulated mutations within the regulatory region of ilvA. By reducing baseline relE expression, these more ‘benign’ mutations lowered circuit burden, which suppressed the accumulation of relE-inactivating mutations, thereby prolonging kill-switch function. Overall, this work demonstrates the utility of sequence entanglement paired with an adaptive laboratory evolution campaign to increase the evolutionary stability of burdensome synthetic circuits.


SUPPLEMENTAL DISCUSSION
Based on our observations in this work, we anticipate that entanglement designs will, in most cases, require post-entanglement optimization of an internal RBS to facilitate translation of the internal gene. In this proof-of-concept example, we achieved toxic levels of RelE using a post-hoc RBS optimization approach whereby strong RBS sites that minimize disruptive amino acid substitutions within IlvA were identified and introduced. Fortuitously, one such RBS site (RBS3) not only improved translation from the embedded relE gene but also enhanced the functionality of IlvA ( Fig. 2A-B and S3). We posit that the entanglement of WT relE within the genetic region encoding the C-terminal regulatory domain of ilvA reduces or alters the autoregulation of IlvA; this phenomenon may also explain the diminished functionality of ilvA/relE STOP compared to ilvA WT in minimal medium without isoleucine ( Fig. 1A and 2A). Since the RBS modifications occur in the alpha-helical "neck" region that separates the catalytic and regulatory domains of IlvA (Fig. 1A), we speculate that the amino acid changes imposed by RBS3 (Fig. S5, ilvA L333V/G334L ) may have locked the N-terminus of IlvA into a catalytically active state. This would allow for the constitutive production of isoleucine and support the enhanced growth that we observed with the RBS3 modification without altering IlvA expression levels ( Fig.  S6B). Confirming this hypothesis will require further biochemical and structural characterization. While we do not expect this fortuitous result to generalize to other entanglement pairs, it does highlight an important design constraint of entanglement: mutations that improve the strength of the internal RBS may impact the fitness of the external entangled gene. Entanglement algorithms that incorporate RBS modifications along with fitness prediction would be highly valuable for future entanglement designs. protegens Pf-5, ilvA and tdcB, a close homolog of ilvA, are both required for isoleucine biosynthesis. Strains were grown in minimal medium with or without the addition of isoleucine. Growth is reported as OD600 over time. Data are shown as the mean of 3 independent replicates.

Fig S3.
Testing the ability of ilvA/relE STOP alleles to rescue isoleucine auxotrophy. Final density of isoleucine auxotroph (ΔilvAΔtdcB) strains with chromosomally integrated antitoxin (PcymR-relB) harboring plasmids carrying the different alleles listed driven under the PrhaBAD promoter. (A) Strains harboring ilvA/relE STOP alleles containing different strength RBSs were grown in minimal medium with addition of either rhamnose (to induce the PrhaBAD -ilvA/relE alleles) or isoleucine. Growth is reported as OD600 after 15 h. (B) Growth curves from data represented in figure 2A and S3A in which growth is reported as OD600 over time. All data in bar graphs are shown as the mean ± SD and data in growth curves are shown as the mean. All data are representative of 3 independent replicates. in which growth is reported as OD600 over time. (A) Strains (ΔilvAΔtdcB, PcymR-relB) harboring PrhaBAD-ilvA/relE alleles with different strength RBSs were grown in rich medium to isolate relE toxicity. Experiments were performed without rhamnose induction of ilvA/relE expression. (B) To rescue growth, the antitoxin was induced by addition of cumate. Data is shown as the mean of 3 independent replicates.

Fig S5. Optimization of internal RBS to improve RelE expression.
The nucleotide sequence (top) and amino acid sequence (bottom) of the internal RBS region within ilvA (red underline), just upstream of the start codon of entangled relE, is listed next to each variant tested. Changes to the nucleotide sequence of ilvA are seen with lowercase letters and the corresponding changes in the amino acid sequence are depicted in bold red font. The predicted translation rate was determined using the Salis Lab RBS calculator (1). The rates are given on a proportional scale ranging from 1 to 100,000+.

Fig S6. Improved RBS modification does not affect threonine deaminase expression. (A)
Growth curve in minimal medium without isoleucine shows that 3xFLAG tagged ilvA constructs are functional for isoleucine biosynthesis. Growth is reported as OD600 over time. (B) Western blot testing whether entanglement and/or RBS3 modification affects expression of threonine deaminase. Strains were grown overnight with rhamnose for induction of the ilvA variants. Growth curve data are shown as the mean and all data are representative of 3 independent replicates.

Fig S7. The C-terminus of ilvA is important for maximal cell fitness. (A) A ΔilvAΔtdcB
PcymR-relB P. protegens strain harboring a vector containing the listed alleles under a rhamnose inducible promoter were grown in minimal medium without isoleucine and with rhamnose to probe the function of ilvA. The ilvA ΔH322-G514 allele contains a C-terminal truncation of ilvA immediately upstream of the internal RBS for relE. Growth is reported as OD600 over time. (B) Competition assay in which ilvA WT , ilvA/relE RBS3 and ilvA ΔH322-G514 were grown in a 1:1 co-culture in minimal medium in the presence or absence of rhamnose (0.001%). Strains were differentially marked with chromosomally integrated antibiotic cassettes, tetracycline and gentamicin. The competitive index was calculated as the CFU ratio of mutant/parent after growth for 48 hours divided by the CFU ratio of the mutant/parent in the initial inoculum. Growth curve data for panel A are shown as the mean of at 3 independent replicates and data for panel B are shown as the mean ± SD of 4 independent replicates. Asterisk(s) directly above data denote comparisons to the ilvA WT : ilvA WT co-culture for each condition. Comparisons were made by one-way ANOVA with Dunnet's post test. *** = P < 0.001, ** = P < 0.01, ns = not significant. RBSs were grown in minimal medium with addition of inducers as listed. Growth is reported as OD 600 after 15 h. (B) Growth curves from data in panel A in which the OD 600 was measured over time for each strain. Data in bar graphs are shown as the mean ± SD and data in growth curves are shown as the mean. All data are representative of 3 independent replicates.

Fig S9. Evolutionary stability of non-entangled relE RBS3 . Independent lineages (ΔilvAΔtdcB
PcymR-relB P. protegens) harboring relE RBS3 (blue lines) were grown in minimal medium with rhamnose (to induce relE RBS3 ) and cumate (to induce antitoxin) in the presence of isoleucine. Each day (~6.6 generations) the cultures were diluted 1:1,000 in fresh medium and plated for CFU on toxin permissive and non-permissive conditions and a survival frequency was calculated. The grey bar indicates toxin escape ratio ≥ 10 -1 (10%). Each line represents data from one of ten independent replicates picked from single colonies. The RBS3 modification was engineered to drive the expression of Prham-relE to a similar level as ilvA/relE RBS3 (Fig. S8).
Escape ratio data for the ilvA/relE RBS3 strain grown with isoleucine (purple lines) are identical to those shown in Fig. 4A and are plotted again here to facilitate comparisons to the relE RBS3 strain.
Fig S10. Growth of isolated colonies from each lineage of the long-term evolutionary stability assay. Single colonies were isolated from each lineage grown without isoleucine on permissive medium during the final passage of the long-term evolutionary stability assay. Each original colony was grown in minimal medium and the OD600 was measured over time (panels A and C) and a bar graph representing OD600 after 15 h of growth was generated (panels B and D). (A, B) To assess isoleucine auxotrophy, minimal medium without isoleucine was supplemented with rhamnose and cumate. (C, D) To assess relE toxicity, minimal medium was supplemented with rhamnose and isoleucine and without cumate. Data in bar graphs are shown as the mean ± SD and data in growth curves are shown as the mean. All data are representative of 3 independent replicates.

Fig S11. Growth of new clean background (CB) strains harboring isolated vectors from each lineage of the long-term evolutionary stability assay.
Vectors were isolated from the colonies of each lineage grown without isoleucine and transformed into a clean genetic background (ΔilvAΔtdcB PcymR-relB) and were grown in (A) minimal medium with isoleucine and rhamnose and without cumate to test relE activity or (B) in minimal medium without isoleucine with addition of rhamnose and cumate to probe ilvA function. Growth curves represent data from figure 5B and 5C. Data are shown as the mean of 3 independent replicates.

Fig S12. Growth ratio of cells that can survive on minimal medium. The indicated strains
were grown overnight in LB medium, washed twice in minimal medium and then serially diluted onto either LB-agar or M9-agar plates without isoleucine. After 48 h of growth, growth ratio was calculated by taking the CFU/mL on M9 plates divided by the CFU/mL on LB plates. No colonies were observed for one replicate for the ∆ilvA∆tdcB strain and this was reported as a limit of detection (one colony) at the lowest dilution plated. Data are shown as the mean ± SD of 4 independent replicates.  WT S1, S12 DP1345 ΔilvA ΔilvA S1 DP1539 ΔilvA ΔtdcB ΔilvA ΔtdcB S1, S12 DP1545 3xFLAG-ilvA ΔilvA ΔtdcB PcymR-relB @glmS pJUMP24-T24-PrhaBAD-3xFLAG-Ec ilvA