Burden-driven feedback control of gene expression

Abstract

Cells use feedback regulation to ensure robust growth despite fluctuating demands for resources and differing environmental conditions. However, the expression of foreign proteins from engineered constructs is an unnatural burden that cells are not adapted for. Here we combined RNA-seq with an in vivo assay to identify the major transcriptional changes that occur in Escherichia coli when inducible synthetic constructs are expressed. We observed that native promoters related to the heat-shock response activated expression rapidly in response to synthetic expression, regardless of the construct. Using these promoters, we built a dCas9-based feedback-regulation system that automatically adjusts the expression of a synthetic construct in response to burden. Cells equipped with this general-use controller maintained their capacity for native gene expression to ensure robust growth and thus outperformed unregulated cells in terms of protein yield in batch production. This engineered feedback is to our knowledge the first example of a universal, burden-based biomolecular control system and is modular, tunable and portable.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Figure 1: Characterization of host–construct interactions in E. coli.
Figure 2: Global transcriptional changes and activation of genomic σ32-regulated promoters in response to synthetic construct expression.
Figure 3: Burden-driven feedback control system.

Accession codes

Primary accessions

Gene Expression Omnibus

References

  1. 1

    Lynch, M. & Marinov, G.K. The bioenergetic costs of a gene. Proc. Natl. Acad. Sci. USA 112, 15690–15695 (2015).

    CAS  Article  Google Scholar 

  2. 2

    Gyorgy, A. et al. Isocost lines describe the cellular economy of genetic circuits. Biophys. J. 109, 639–646 (2015).

    CAS  Article  Google Scholar 

  3. 3

    Qian, Y., Huang, H.-H., Jiménez, J.I. & Del Vecchio, D. Resource competition shapes the response of genetic circuits. ACS Synth. Biol. 6, 1263–1272 (2017).

    CAS  Article  Google Scholar 

  4. 4

    Weiße, A.Y., Oyarzún, D.A., Danos, V. & Swain, P.S. Mechanistic links between cellular trade-offs, gene expression, and growth. Proc. Natl. Acad. Sci. USA 112, E1038–E1047 (2015).

    Article  Google Scholar 

  5. 5

    Kurland, C.G. & Dong, H. Bacterial growth inhibition by overproduction of protein. Mol. Microbiol. 21, 1–4 (1996).

    CAS  Article  Google Scholar 

  6. 6

    Sleight, S.C. & Sauro, H.M. Visualization of evolutionary stability dynamics and competitive fitness of Escherichia coli engineered with randomized multigene circuits. ACS Synth. Biol. 2, 519–528 (2013).

    CAS  Article  Google Scholar 

  7. 7

    Cardinale, S. & Arkin, A.P. Contextualizing context for synthetic biology—identifying causes of failure of synthetic biological systems. Biotechnol. J. 7, 856–866 (2012).

    CAS  Article  Google Scholar 

  8. 8

    Moser, F. et al. Genetic circuit performance under conditions relevant for industrial bioreactors. ACS Synth. Biol. 1, 555–564 (2012).

    CAS  Article  Google Scholar 

  9. 9

    Tan, C., Marguet, P. & You, L. Emergent bistability by a growth-modulating positive feedback circuit. Nat. Chem. Biol. 5, 842–848 (2009).

    CAS  Article  Google Scholar 

  10. 10

    Borkowski, O., Ceroni, F., Stan, G.-B. & Ellis, T. Overloaded and stressed: whole-cell considerations for bacterial synthetic biology. Curr. Opin. Microbiol. 33, 123–130 (2016).

    CAS  Article  Google Scholar 

  11. 11

    Nielsen, A.A.K. et al. Genetic circuit design automation. Science 352, aac7341 (2016).

    Article  Google Scholar 

  12. 12

    Cameron, D.E. & Collins, J.J. Tunable protein degradation in bacteria. Nat. Biotechnol. 32, 1276–1281 (2014).

    CAS  Article  Google Scholar 

  13. 13

    Segall-Shapiro, T.H., Meyer, A.J., Ellington, A.D., Sontag, E.D. & Voigt, C.A.A. A 'resource allocator' for transcription based on a highly fragmented T7 RNA polymerase. Mol. Syst. Biol. 10, 742 (2014).

    Article  Google Scholar 

  14. 14

    Pasini, M. et al. Using promoter libraries to reduce metabolic burden due to plasmid-encoded proteins in recombinant Escherichia coli. N. Biotechnol. 33, 78–90 (2016).

    CAS  Article  Google Scholar 

  15. 15

    Gorochowski, T.E., Avcilar-Kucukgoze, I., Bovenberg, R.A.L., Roubos, J.A. & Ignatova, Z. A minimal model of ribosome allocation dynamics captures trade-offs in expression between endogenous and synthetic genes. ACS Synth. Biol. 5, 710–720 (2016).

    CAS  Article  Google Scholar 

  16. 16

    Carrera, J., Rodrigo, G., Singh, V., Kirov, B. & Jaramillo, A. Empirical model and in vivo characterization of the bacterial response to synthetic gene expression show that ribosome allocation limits growth rate. Biotechnol. J. 6, 773–783 (2011).

    CAS  Article  Google Scholar 

  17. 17

    Ceroni, F., Algar, R., Stan, G.-B. & Ellis, T. Quantifying cellular capacity identifies gene expression designs with reduced burden. Nat. Methods 12, 415–418 (2015).

    CAS  Article  Google Scholar 

  18. 18

    Shachrai, I., Zaslaver, A., Alon, U. & Dekel, E. Cost of unneeded proteins in E. coli is reduced after several generations in exponential growth. Mol. Cell 38, 758–767 (2010).

    CAS  Article  Google Scholar 

  19. 19

    Wang, Z., Gerstein, M. & Snyder, M. RNA-Seq: a revolutionary tool for transcriptomics. Nat. Rev. Genet. 10, 57–63 (2009).

    CAS  Article  Google Scholar 

  20. 20

    Houser, J.R. et al. Controlled measurement and comparative analysis of cellular components in E. coli reveals broad regulatory changes in response to glucose starvation. PLOS Comput. Biol. 11, e1004400 (2015).

    Article  Google Scholar 

  21. 21

    Nonaka, G., Blankschien, M., Herman, C., Gross, C.A. & Rhodius, V.A. Regulon and promoter analysis of the E. coli heat-shock factor, σ32, reveals a multifaceted cellular response to heat stress. Genes Dev. 20, 1776–1789 (2006).

    CAS  Article  Google Scholar 

  22. 22

    Guisbert, E., Herman, C., Lu, C.Z. & Gross, C.A. A chaperone network controls the heat shock response in E. coli. Genes Dev. 18, 2812–2821 (2004).

    CAS  Article  Google Scholar 

  23. 23

    Farasat, I. & Salis, H.M. A biophysical model of CRISPR/Cas9 activity for rational design of genome editing and gene regulation. PLOS Comput. Biol. 12, e1004724 (2016).

    Article  Google Scholar 

  24. 24

    Jiang, L. et al. Synthetic spike-in standards for RNA-seq experiments. Genome Res. 21, 1543–1551 (2011).

    CAS  Article  Google Scholar 

  25. 25

    Arkin, A.P. & Cambray, G. Massive phenotypic measurements reveal complex physiological consequences of differential translation efficacies. bioRxiv Preprint at https://www.biorxiv.org/content/early/2017/10/25/209098 (2017).

    Google Scholar 

  26. 26

    Brar, G.A. & Weissman, J.S. Ribosome profiling reveals the what, when, where and how of protein synthesis. Nat. Rev. Mol. Cell Biol. 16, 651–664 (2015).

    CAS  Article  Google Scholar 

  27. 27

    Lohman, B.K., Weber, J.N. & Bolnick, D.I. Evaluation of TagSeq, a reliable low-cost alternative for RNAseq. Mol. Ecol. Resour. 16, 1315–1321 (2016).

    CAS  Article  Google Scholar 

  28. 28

    Zhang, X. et al. Heat-shock response transcriptional program enables high-yield and high-quality recombinant protein production in Escherichia coli. ACS Chem. Biol. 9, 1945–1949 (2014).

    CAS  Article  Google Scholar 

  29. 29

    El-Samad, H., Kurata, H., Doyle, J.C., Gross, C.A. & Khammash, M. Surviving heat shock: control strategies for robustness and performance. Proc. Natl. Acad. Sci. USA 102, 2736–2741 (2005).

    CAS  Article  Google Scholar 

  30. 30

    Kurata, H. et al. Module-based analysis of robustness tradeoffs in the heat shock response system. PLOS Comput. Biol. 2, e59 (2006).

    Article  Google Scholar 

  31. 31

    Dragosits, M., Nicklas, D. & Tagkopoulos, I. A synthetic biology approach to self-regulatory recombinant protein production in Escherichia coli. J. Biol. Eng. 6, 2 (2012).

    CAS  Article  Google Scholar 

  32. 32

    Gupta, A., Reizman, I.M., Reisch, C.R. & Prather, K.L.J. Dynamic regulation of metabolic flux in engineered bacteria using a pathway-independent quorum-sensing circuit. Nat. Biotechnol. 35, 273–279 (2017).

    CAS  Article  Google Scholar 

  33. 33

    Nakahigashi, K., Yanagi, H. & Yura, T. Regulatory conservation and divergence of 32 homologs from Pseudomonas aeruginosa, and Agrobacterium tumefaciens. Microbiology 180, 2402–2408 (1998).

    CAS  Google Scholar 

  34. 34

    Casini, A. et al. R2oDNA designer: computational design of biologically neutral synthetic DNA sequences. ACS Synth. Biol. 3, 525–528 (2014).

    CAS  Article  Google Scholar 

  35. 35

    Gertz, J. et al. Transposase mediated construction of RNA-seq libraries. Genome Res. 22, 134–141 (2012).

    CAS  Article  Google Scholar 

  36. 36

    He, S. et al. Validation of two ribosomal RNA removal methods for microbial metatranscriptomics. Nat. Methods 7, 807–812 (2010).

    CAS  Article  Google Scholar 

  37. 37

    Gorochowski, T.E. et al. Genetic circuit characterization and debugging using RNA-seq. Mol. Syst. Biol. 13, 952 (2017).

    Article  Google Scholar 

  38. 38

    Myers, C.J. et al. A standard-enabled workflow for synthetic biology. Biochem. Soc. Trans. 45, 793–803 (2017).

    CAS  Article  Google Scholar 

  39. 39

    Der, B.S. et al. DNAplotlib: programmable visualization of genetic designs and associated data. ACS Synth. Biol. 6, 1115–1119 (2017).

    CAS  Article  Google Scholar 

  40. 40

    Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).

    Article  Google Scholar 

  41. 41

    Keseler, I.M. et al. The EcoCyc database: reflecting new knowledge about Escherichia coli K-12. Nucleic Acids Res. 45, D543–D550 (2017).

    CAS  Article  Google Scholar 

  42. 42

    Edgar, R., Domrachev, M. & Lash, A.E. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res 30, 207–210 (2002).

    CAS  Article  Google Scholar 

Download references

Acknowledgements

This work was supported by EPSRC (grant EP/J021849/1 to F.C., G.-B.S. and T.E.; grant EP/M002306/1 to A.B., O.B. and C.G.; grant EP/P009352/1 to G.-B.S.; Fellowship EP/M002187/1 to G.-B.S.; Fellowship EP/M002306/1 to T.E.), BBSRC/EPSRC SBRC BrisSynBio (grant BB/L01386X/1 to T.E.G.), the Royal Society (Fellowship UF160357 to T.E.G.), the NIHR Imperial Biomedical Research Centre (Y.N.L.) and BBSRC (grant BB/K006290/1 to A.R.A.). F.C. acknowledges the support of the Imperial College London Junior Research Fellowship Scheme. All authors thank G. Cambray for useful discussions.

Author information

Affiliations

Authors

Contributions

F.C., G.-B.S. and T.E. designed the research; F.C., A.B., C.G. and A.R.A. performed the experiments; F.C., S.F., T.E.G., Y.N.L., G.-B.S. and T.E. analyzed the data; F.C., S.F., T.E.G., O.B., G.-B.S. and T.E. wrote the paper.

Corresponding authors

Correspondence to Guy-Bart Stan or Tom Ellis.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Integrated supplementary information

Supplementary Figure 1 Burden assay for cells transformed with the synthetic constructs 1 h after induction of gene expression.

Left and middle panels show the growth rate in DH10B and MG1655, respectively. Grey and pink bars represent uninduced and induced samples, respectively. Right panel shows GFP production rate in MG1655. Grey and green bars represent uninduced and induced samples, respectively. Darker grey bars represent the uninduced samples; lighter grey bars represent the empty plasmids while the lighter top bars represent the growth rate in the untransformed strains. Dots show single repeats while bars show the mean value for two independent repeats.

Supplementary Figure 2 Correlations between RNA-seq data and the in vivo burden assay.

Left panel shows the relationship between the total RNA harvested during RNAseq sample prep and the total cells grown for each sample in the plate-based burden assay done in parallel. Correlation is given as R2 - the square of the Pearson product moment correlation coefficient. Cell concentration per well was determined by OD600 measurement at each time point (15 min and 60 min post-induction). RNA concentrations from extractions into the same volume were determined by spectrographic measurement with Nanodrop. Data represent values from three independent repeats. Right panel shows the relationship between RNA-seq GFP FPKM (x-axis) and in vivo GFP production rate per cell (y-axis). Correlation is given as R2 - the square of the Pearson product moment correlation coefficient. Data are from 15 and 60 min post-induction. Control samples are the strains (DH = DH10B, MG = MG1655) with the three empty plasmids (pSB1C3, pLys and pD864). Burden samples are the strains with the four synthetic constructs (pSB1C3-Lux, pSB1C3-H3, pLys-M1 and pD864-LacZ). Data show the mean values for each condition.

Supplementary Figure 3 Burden imposed by the four synthetic constructs.

(a) Transcription profiles of the synthetic constructs pD864-LacZ, pLys-M1 and pSB1C3-Lux in DH10B cells 15 (top) and 60 (bottom) min post-induction. Traces show the normalized reads mapping to the plasmid constructs. Solid grey regions show average, and transparent red regions span from minimum to maximum of transcription profiles generated from three biological replicates. Genetic designs below each profile are drawn using SBOLv notation and show the precise location of each genetic part. For pD864-LacZ, only reads uniquely mapping to the plasmid lacZ gene (and not the chromosomal copy) are shown. These are used to estimate the level of plasmid lacZ expression (blue shaded regions, see Online Methods section). (b) Transcriptional induction of the four constructs represented by mean FPKM values in MG1655 at 1h post induction. Green bars are data for the calculated GFP FPKM and colour intensity relates to the time point; grey bars are data for the FPKM corresponding to the construct gene; * denotes empty plasmid controls. Dots and bars show individual measurements and their mean, respectively. Error bars show the standard deviation of three independent repeats. (c) Percentage of available capacity 1h post induction. In the left panels, plots show GFP mRNA FPKM in construct samples compared to the pD864 control. In the right panels, plots show the same comparison but using GFP fluorescence production rate data from the plate-based burden assay (Figure 1c). Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats.

Supplementary Figure 4 Transcriptome data analysis.

(a) Venn diagrams of differentially expressed genes in MG1655 E. coli 1h post-induction. The set of significantly differentially expressed native genes (alpha < 0.05) was identified separately for each of three synthetic gene circuits comparing cells transformed with the synthetic circuits to cells transformed with the corresponding empty plasmids. Up-regulated and down-regulated genes for each sample are displayed in black and red, respectively; (b) List of top ten differentially expressed genes (black are up-regulated, red are down-regulated). Tables report the top ten differentially expressed genes for each construct and the associated regulatory sigma factors. Bold names indicate genes differentially expressed also at 15 min post-induction. Left columns show the differentially expressed genes that are shared among all 3 constructs; (c) Volcano plots. Volcano plots show genes that display large changes and are also statistically significant (DeSeq2, n=3). Symbol colors correspond to the constructs colors displayed in (a) with pSB1C3-Lux, pD864-LacZ and pSB1C3-H3 displayed in yellow, blue and red, respectively. (d) Transcription profiles of the dnaKJ, htpG, ibpAB and groSL regulons and the GFP monitor cassette from DH10B cells with induced synthetic construct pD864-LacZ and pSB1C3-H3. Top profiles are for strains containing only the empty plasmid 15 min post-induction, middle and bottom profiles are for strains containing synthetic constructs 15 and 60 min post-induction, respectively. Solid grey regions show average and transparent red regions span the min to max of transcription profiles generated from three biological replicates. Genetic designs below each profile are drawn using SBOLv notation.

Supplementary Figure 5 Global transcriptional changes in response to burden 15 min after induction of synthetic construct expression.

The set of significantly differentially expressed native genes (alpha < 0.05) was identified in DH10B (top panels) and MG1655 (bottom panels) cells separately for each of the three synthetic gene circuits pSB1C3-Lux, pSB1C3-H3 and pD864-LacZ, comparing cells transformed with the synthetic circuits to cells transformed with the corresponding empty plasmids. (a) Venn diagrams of differentially expressed genes. Venn diagrams report the number of up-regulated and down-regulated genes for each sample in DH10B and MG1655, which are displayed in black and red, respectively; (b) List of top ten differentially expressed genes (black are up-regulated, red are down-regulated). Tables report the top ten differentially expressed genes for each construct and the associated regulatory sigma factors. Left columns show the differentially expressed genes that are shared among all 3 constructs; (c) Volcano plots. Volcano plots show genes that display large changes and are also statistically significant (DeSeq2, n=3). Symbol colors correspond to the constructs colours displayed in (a) with pSB1C3-Lux, pD864-LacZ and pSB1C3-H3 displayed in yellow, blue and red, respectively.

Supplementary Figure 6 Global transcriptional changes in response to burden 15 and 60 min post-induction for all four synthetic constructs.

The set of significantly differentially expressed native genes (alpha < 0.05) at 15 min (panels a, b and c) and 60 min (panels d, e and f) after induction of gene expression was identified separately for each of four synthetic gene circuits pSB1C3-Lux, pSB1C3-H3, pLys-M1 and pD864-LacZ, comparing cells transformed with the synthetic circuits to cells transformed with the corresponding empty plasmids. (a, d) Venn diagrams of differentially expressed genes. Venn diagrams report the number of up-regulated and down-regulated genes for each sample in DH10B (top) and MG1655 (bottom), which are displayed in black and red, respectively; (b, e) List of top ten differentially expressed genes (black are up-regulated, red are down-regulated). Tables report the top ten differentially expressed genes for each construct and the associated regulatory sigma factors in DH10B (top) and MG1655 (bottom). (c, f) Volcano plots. Volcano plots show genes that display large changes and are also statistically significant (DeSeq2, n=3) in DH10B (top) and MG1655 (bottom). Symbol colours correspond to the constructs colours displayed in (a) with pSB1C3-Lux, pD864-LacZ, pSB1C3-H3 and pLys-M1 represented in yellow, blue, pink, and cyan respectively.

Supplementary Figure 7 Characterization of burden-dependent activation of plasmid-based σ32-regulated promoters.

The in vivo response to burden from the htpG1, htpG2, groSL and ibpAB promoters. All four promoters were assessed for their ability to induce GFP expression from a high-copy plasmid when co-transformed with a construct expressing VioB-mCherry upon arabinose induction. Data show the GFP production rate per cell from the four promoters 2 hours (top panel) and 1 hour (middle panel) post-induction and VioB-mCherry expression 1 hour post induction (bottom panel) in DH10B cells. Dark and light green bars represent uninduced (no arabinose) and induced samples (arabinose), respectively. Dots show single measurements and bar show the mean value of two independent measurements.

Supplementary Figure 8 Biomolecular feedback controller.

(a) Modular structure of the feedback system. The feedback system was placed on a medium copy number plasmid with the ampicillin resistance marker. The feedback is composed of a dCas9 cassette with constitutive expression from a synthetic promoter and of a downstream burden-responsive cassette. In the burden-responsive cassette the burden-responsive htpG1 promoter controls the production of a sgRNA, in this case targeting the pBAD core promoter. The system was conceived with a modular structure where the htpG1 promoter and the gRNA can be easily replaced using rare restriction enzymes flanking each module. (b) Burden imposed by the molecular feedback controller on DH10B cells. Top panels show the growth rates (left panel) and GFP production rates (right panel) 1 hour post induction for uninduced DH10B cells (with integrated capacity monitor) when no construct is present (DH10B), and uninduced cells containing both pSB1C3-H2 and the feedback system plasmids (PdCas9-X), or uninduced cells containing just pSB1C3-H2 (H2). The burden imposed by the feedback system when promoters of different strength are present upstream of dCas9 is shown (PdCas9-1 to 6). Bottom panels show growth rates, as well as GFP and mCherry production rates when the same constructs are induced with arabinose. H2 is a previously-designed and tested construct, similar to pSB1C3-H3, but with a stronger RBS and so imposing even higher levels of burden. For a full description of H2, please see Ceroni et al, Nature Methods 2015. Dots show single measurements and bar show the mean value. Error bars show the standard deviation of three biologically independent repeats.

Supplementary Figure 9 Test of the robustness of the feedback system at different temperatures.

GFP (top) and VioB-mCherry (bottom) production rate per cell are shown when different levels of arabinose are used to trigger VioB-mCherry production in DH10B cells at 37°C and 30°C. The performance of the feedback system in presence of construct pSB1C3-H2 (Closed-Loop) was compared to an ‘Open-Loop’ version where the sgRNA targeting pBAD had been replaced by a random sgRNA with no affinity for any promoter. For both strains VioB-mCherry output from pSB1C3-H2 (lower panels) and cellular capacity (top panels) were characterised when different levels of arabinose were used to induce construct expression. Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats.

Supplementary Figure 10 Portability of the feedback controller in multiple E. coli strains.

(a) Comparison between the feedback system and the unregulated strain. The functionality of the controller was tested in BL21-DE3 (left panels) and MG1655 (right panels). The performance of the feedback system in presence of construct pSB1C3-H2 (Closed-Loop) was compared to the strain where only pSB1C3-H2 is present (H2). For both strains VioB-mCherry output from pSB1C3-H2 (lower panels) and cellular capacity (top panels) were characterised when different levels of arabinose are used to induce construct expression. Since BL21-DE3 cells do not bear the capacity monitor, growth rate 3 hours post-induction was used as a proxy for cellular capacity instead of the GFP production rate per cell. Error bars show standard deviation of three independent repeats. (b) Comparison between the feedback system and the open loop strain. The performance of the feedback system in presence of construct pSB1C3-H2 (Closed-Loop) was compared to an ‘Open-Loop’ version where the sgRNA targeting pBAD had been replaced by a random sgRNA with no affinity for any promoter. Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats.

Supplementary Figure 11 The biomolecular feedback controller retargeted to a different synthetic construct.

Growth rate (left), GFP expression rate (middle) and total luminescence per cell (right) are shown for a construct where the firefly luciferase operon is expressed under the control of the lacP promoter (lacP-Luc) in DH10B E. coli. In cells where the lacP-targeting closed-loop feedback controller is alongside the luciferase operon plasmid (FB), luminescence is decreased while growth rate and GFP production rate are improved. Sequences of the lacP core promoter and sgRNA targeting this are both shown in Table S7. Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats.

Supplementary Figure 12 Presence of the feedback controller confers robustness to temperature change.

(a) Top panels show the growth rate, and GFP and VioB-mCherry production rates 1 hour post induction at 37ºC in DH10B cells. When pSB1C3-H2 with the closed-loop feedback control plasmid (FB) and pSB1C3-H2 alone (H2) strains are induced with 0.5% and 0.002% arabinose, respectively, they both behave very similarly 1 hour post-induction in terms of growth rate, and GFP and VioB-mCherry production rates. Lower panels show the growth rate, and GFP and VioB-mCherry production rates 1 hour post induction at 30ºC. At 30ºC, H2 now produces significantly more VioB-mCherry, but is accompanied by a decrease in cellular capacity and growth rate, while FB maintains VioB-mCherry expression and keeps the desired high expression capacity and growth. Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats. (b) The biomolecular feedback was tested in DH10B at 37°C and 30°C and MG1655 cells at 37°C with integrated capacity monitor (left panels) and in BL21-DE3 cells at 37°C (right panels). GFP (top) and VioB-mCherry (bottom) production rate per cell are shown for DH10B and MG1655 when 0% and 1% of arabinose are added. For BL21-DE3, growth rate (top) and VioB-mCherry rate (bottom) are shown instead. Dots show single repeats while bars show the mean value. Error bars show standard deviation of three independent repeats.

Supplementary Figure 13 Presence of the feedback controller confers robustness to varying concentrations of inducer.

Top and bottom panels show VioB-mCherry and GFP production rates 1 hour post induction at different arabinose concentrations, respectively, in DH10B cells. While pSB1C3-H2 alone (H2) is highly sensitive to increasing concentrations of arabinose, pSB1C3-H2 with the feedback controller plasmid (FB) is capable of maintaining the GFP production rate per cell (i.e. the host capacity) and the VioB-mCherry production rates per cell within a much narrower range, thereby demonstrating robustness of the feedback system to inducer fluctuations. Dots represent the mean and the error bars indicate the maximum and minimum values of the triplicate.

Supplementary Figure 14 Burden-driven feedback control in a 24-h batch time course.

(a) Constructs FB, FB2 and FB5 from Figure 3c were selected for a 24 hour batch growth time course after addition of 1% arabinose to the cultures containing the feedback system and pSB1C3-H2 construct (H2). The culture density (left panel), total VioB-mCherry production per cell (middle panel) and total VioB-mCherry yield in growth flasks are compared for time points from 16 to 24 hours. Plots show how increased growth due to alleviation of burden can compensate the decreased VioB-mCherry production per cell, leading to higher total protein yields over time. Interestingly, looking at Figure S15 at 20 hours it is possible to compare the mCherry per cell values measured using the plate reader with the corresponding values obtained through flow cytometry measurement. Both plate reader and flow cytometry assays confirm H2 (no feedback control) and FB2 as the constructs producing the most VioB-mCherry per cell, while FB5 produces a lower amount of VioB-mCherry per cell, and FB (the closed-loop feedback controller) produces the least amount per cell. (b) Earlier time points of a 24 hour time-course experiment are shown. Data correspond to a time-course from 0 to 12 hours for culture density (left) and total VioB-mCherry yield (right) in growth flasks. To avoid artefacts due to measurements and normalisation with very low ODs, this experiment was started at a higher initial OD.

Supplementary Figure 15 Single-cell 24-h time-course experiment of the burden-driven feedback controls.

Graphs show single-cell flow cytometry measurements of the population corresponding to the plate reader data shown in Figure 3d. The geometric mean values of sfGFP and VioB-mCherry expression per cell are given next to the histogram distributions. Plots confirm that no escape mutant arose during the batch culture dynamic experiment. The experiment was repeated three times with similar results.

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–15, Supplementary Tables 1–7 and Supplementary Notes 1–3

Life Sciences Reporting Summary

Supplementary Data 1

Analyzed DH10B RNA-seq data

Supplementary Data 2

Analyzed MG1655 RNA-seq data

Supplementary Data 3

Source data for supplementary figures

Source data

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Ceroni, F., Boo, A., Furini, S. et al. Burden-driven feedback control of gene expression. Nat Methods 15, 387–393 (2018). https://doi.org/10.1038/nmeth.4635

Download citation

Further reading

Search

Sign up for the Nature Briefing newsletter for a daily update on COVID-19 science.
Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing