1、LETTERSA fast, robust and tunable synthetic gene oscillatorJesse Stricker1*, Scott Cookson1*, Matthew R. Bennett1,2*, William H. Mather1, Lev S. Tsimring2almost every cell exhibitedlarge-amplitude fluorescence oscillations throughout observationruns. The oscillatory period can be tuned by altering i
2、nducerlevels, temperature and the media source. Computational model-ling demonstrates that the key design principle for constructing arobust oscillator is a time delay in the negative feedback loop,which can mechanistically arise from the cascade of cellular pro-cesses involved in forming a function
3、al transcription factor. Thepositive feedback loop increases the robustness of the oscillationsand allows for greater tunability. Examination of our refinedmodel suggested the existence of a simplified oscillator designwithout positive feedback, and we construct an oscillator strainconfirming this c
4、omputational prediction.The synthetic gene oscillator is based on a previously reportedtheoretical design1and was constructed using E. coli components(Fig. 1a). The hybrid promoter (plac/ara-1; ref. 14) is composed ofthe activation operator site from the araBAD promoter placed inits normal location
5、relative to the transcription start site, and repres-sion operator sites from the lacZYA promoter placed both upstreamand immediately downstream of the transcription start site. It isactivated by the AraC protein in the presence of arabinose andrepressed by the LacI protein in the absence of isoprop
6、ylb-D-1-thio-galactopyranoside (IPTG). We placed the araC, lacI and yemGFP(monomeric yeast-enhanced green fluorescent protein) genes underthe control of three identical copies of plac/ara-1to form three co-regulated transcription modules (Supplementary Infor-mation). Hence, activation of the promote
7、rs by the addition of ara-binose and IPTG to the medium results in transcription of eachcomponent of the circuit, and increased production of AraC in thepresence of arabinose results in a positive feedback loop thatincreases promoter activity. However, the concurrent increase inproduction of LacI re
8、sults in a linked negative feedback loop thatdecreases promoter activity, and the differential activity of the twofeedback loops can drive oscillatory behaviour1,13.The oscillator cells (denoted JS011) exhibited ubiquitous fluor-escence oscillations over the entire run time of each experiment (atlea
9、st 4h). For example, at 0.7% arabinose and 2mM IPTG, more1Department of Bioengineering, University of California, San Diego, La Jolla, California 92093, USA.2Institute for Nonlinear Science, University of California, San Diego, La Jolla,California 92093, USA.*These authors contributed equally to thi
10、s work.aceg0 120 18012048Time (min)Fluorescence (a.u.)lacIyemGFParaC+ Arabinose IPTGb60dhTime (min)f0 120 180600 120 180600 120 18060 0 120 18060012018060012018060Figure 1 | Oscillations in the dual-feedback circuit. a, Network diagram ofthedual-feedbackoscillator.Ahybridpromoterplac/ara-1drivestran
11、scriptionofaraC andlacI,formingpositiveandnegativefeedbackloops.b,Single-cellfluorescence trajectories induced with 0.7% arabinose and 2mM IPTG.Points represent experimental fluorescence values, and solid curves aresmoothed by a SavitskyGolay filter (for unsmoothed trajectories, seeSupplementaryFig.
12、3).The trajectoryinredcorrespondstothedensitymapabove the graph. Density maps for trajectories in grey are shown in g. a.u.,arbitrary units. ch, Single-cell density map trajectories for various IPTGconditions (c, 0mM IPTG; d, 0.25mM; e, 0.5mM; f, 1mM; g, 2mM;h, 5mM).Vol 456|27 November 2008|doi:10.1
13、038/nature073895162008 Macmillan Publishers Limited. All rights reservedthan 99% of the cells showed oscillations with a period of approxi-mately40min(Fig.1b,g,SupplementaryTable1andSupplementaryMovie 1). The highly dynamic nature of the oscillator components isshown by the rapid decay of green fluo
14、rescent protein (GFP) signal,which drops from peak to trough in less than 10min (Fig. 1b). Theoscillatoryphasewasheritablebetweendaughtercells,whichresultedinsynchronizedoscillationsinareasofthemicrocolonyderivedfroma common cell. This synchrony was limited to a few periods, pre-sumablyowingtooscill
15、atoryphasediffusion.Weusedamicrofluidicdevice with a laminar boundary switch upstream of the growthchamber to investigate the initiation of synchronized oscillations(Supplementary Fig. 2c, d). Cells grown in the absence of inducerinitiated oscillations in a synchronous manner on the addition ofinduc
16、er (Supplementary Movie 10), which suggested the possibilityof using flow cytometry to characterize the oscillator further. Flowcytometry of samples continuously collected from a culture in loga-rithmic growth that had been induced with 0.7% arabinose and2mM IPTG showed oscillations in mean cell flu
17、orescence(Supplementary Fig. 8). Induction of oscillation was very quick (lessthan 5min) and initially well-synchronized. The amplitude of thesebulk oscillations decayed as the experiment progressed, as expectedfrom the desynchronization of individual cells in the colony(Supplementary Information).
18、However, the period obtained fromthe flow cytometry method (green data points in all figures) com-pared favourably to that obtained from single cells using microscopy(red data points in all figures).The oscillator was extremely robust over an extensive range ofinducer conditions and temperatures. At
19、 0.7% arabinose and 37uC,almost every observed cell oscillated (Supplementary Table 1) at allIPTG concentrations examined (Fig. 1bh and SupplementaryMovies 18). Varying the IPTG concentration allowed for the tuningof the oscillator period (Fig. 2a), particularly at low IPTG concentra-tions. The peri
20、od decreased at high IPTG concentrations, and sub-sequent characterization of the promoter revealed that this non-monotonic behaviour is probably caused by IPTG interference withAraC activation15(Supplementary Information). The cell doublingtime on the microfluidic device remained largely steady bet
21、weenexperiments,rangingfrom22.3minto27.6minat37uCandshowinglittle correlation to IPTG concentration (R250.132). Individual cellfluorescence trajectories showed a gradual increase in oscillatoryperiod as the cells were imaged on the microfluidic device (Supple-mentary Fig. 4). This increase was not s
22、een in doubling times, imply-ing that the cells were not experiencing nutritional difficultiesor environmental stress that might cause an alteration in oscillatorbehaviour.To explore further the robustness of the oscillator, we investigatedtheeffectofvaryingarabinose,temperatureandthemediasource.Ata
23、fixed value of 2mM IPTG and at 37uC, the oscillatory period can betunedfrom13minto58minbyvaryingthearabinoselevelfrom0.1%to 3.0% (Fig. 2b). Cells grown in the absence of arabinose did notexpress measurable levels of GFP in single-cell microscopy or flowcytometryexperiments,andhighlevelsofarabinosese
24、emedtosaturatethe system. We observed sustained oscillations at a range of tempera-tures from 25uCto37uC, with a decreasing period as a function oftemperature (Fig. 2c). The cell doubling time also decreased withtemperature, as expected, and the oscillatory period increased mono-tonically with cell
25、doubling time (Fig. 2d). The oscillator also func-tioned in minimal A medium with 2gl21glucose (Fig. 2c, d).Althoughthe celldoublingtimeinminimalmediumwas significantlylongerthaninLB-Millerformulationlysogenybroth(LB)(8090minversus 2224min at 37uC), the period in the minimal medium wasvery similar t
26、o that in LB (Fig. 2c, d). This result, together with thestrongdependenceoftheperiodonIPTGandarabinoseconcentration(at constant cellular doubling times), demonstrates that the syntheticoscillatorisnot stronglycoupledtothe cellcycle.The similardepend-enceofthe periodandthedoublingtimeonthetemperature
27、seemstobeduetothethermodynamicchangeoftherateconstantsaffectingallcellular processes.Theoscillatorwasconstructedaccordingtodesignprinciplesdeter-mined from previous theoretical work1. However, we found that thisoriginalmodel failed to describetwoimportantaspectsof theexperi-ments. First, the model c
28、ould not describe the observed functionaldependenceoftheperiodoninducerlevels.Second,andperhapsmostimportantly, because careful parameter tuning was necessary foroscillationsintheoriginalmodel,itwasnotabletodescribetherobustbehaviourdemonstratedintheexperiments.Thissuggeststhatonlyasmall region of i
29、nducer space should support oscillations, in contrastto the robust behaviour demonstrated in the experiments. Theseshortcomings forced a re-evaluation of the derivation of the oscilla-tor equations, and led to a new computational model that moreaccurately described the experimental observations. The
30、 new modelincorporates the same coupled positive and negative feedback archi-tecture, but includes details that were omitted from the previousmodel.Inparticular,wefoundthatdirectlymodellingprocessessuchas proteinDNA binding, multimerization, translation, DNA loop-ing, enzymatic degradation and prote
31、in folding greatly increased theaccuracyofthemodel.Theresultisacomputationalmodelthatisveryrobust to parameter variations and correctly describes the dynamicsof the oscillator for a large range of IPTG and arabinose concentra-tions (see Box 1 and Supplementary Information).In examining our refined m
32、odel, we discovered another region inparameter space that would support oscillatory behaviour. Ourmodel predicted that a constantly activated system with repressioncontrolled by a negative feedback loop could produce oscillations inthe absence ofpositive feedback (SupplementaryFig.19). It hasbeenpro
33、posed that negative feedback gene networks can oscillate as longas there is delay in the feedback16,17, and, although there is no explicitdelay in our model, the intermediate steps of translation, proteinfolding and multimerization of LacI provide an effective form ofdelay18that is sufficient to sup
34、port oscillations. We constructed thissystem (denoted JS013) in E. coli using a hybrid promoter, pLlacO-1(ref.14),thatisactivatedintheabsenceofLacI(orpresenceofIPTG)to drive both lacI and yemGFP expression (Fig. 3a). We observedoscillations in these cells when examined by single-cell microscopyunder
35、 inducing conditions (Fig. 3b, Supplementary Fig. 5 andSupplementary Movie 11). These oscillations were not as distinctor regular as in the dual-feedback oscillator, and they did not alwaysa0Oscillatory period (min)60010IPTG (mM)302037 C (G = 2224 min)0.7% arabinose bcd060012001Arabinose (%)32012020
36、Temperature (C)4030 0Cell doubling period (min)1206037 C (G = 2224 min)2 mM IPTG0.7% arabinose2 mM IPTG0.7% arabinose2 mM IPTGOscillatory period (min)Figure 2 | Robust oscillations. ac, Oscillatory periods on transects with0.7% arabinose and varying IPTG (a), 2mM IPTG and varying arabinose(b), or 0.
37、7% arabinose, 2mM IPTG, and varying temperature (c). Meanperiods from single-cell microscopy (red diamonds, mean6s.d.) or flowcytometry (green circles) are shown. Black curves are trend lines in a andb, or represent the theoretical prediction based on reference values at 30uCin c (see Supplementary
38、Information). Samples grown in minimal mediumrather than LB are indicated by crosses. G represents the cell doublingperiod.d,Oscillatoryperiodandcelldivisiontimeincreasemonotonicallyasthe growth temperature decreases. Symbols are as described above, and theblack line is a linear regression of sample
39、s grown in LB.NATURE|Vol 456|27 November 2008 LETTERS5172008 Macmillan Publishers Limited. All rights reservedreturntoadimstate,consistentwiththepredictionsofthecomputa-tional model. Furthermore, the period was largely unaffected byIPTG concentration (varying less than 5% over three experimentalruns
40、 from 0.6mM to 20mM IPTG), suggesting that the addition ofthe positive feedback loop serves the dual role of regularizing oscilla-tions and allowing tunability of the period (see SupplementaryInformation).In the context of synthetic biology, our findings indicate that cau-tion must be exercised when
41、 making simplifying assumptions in thedesign of engineered gene circuits. We found that a full model of thesystem that takes into account intermediate steps such as multimeri-zation,translation,protein foldingandDNA loopingis essential.Thereason for this lies not only in the timescales of the system
42、 but also inthe sequential timing of events. Because the intermediate steps in theproductionoffunctionalproteintaketime,theirintroductionintothemodel creates an important form of delay1820. We found that thiseffective delay greatly increases the robustness of our model. Forinstance, oscillatory acti
43、vity in the model is only somewhat sensitiveto the values chosen for system parameters (SupplementaryInformation), implying that nearly all cells should oscillate(Supplementary Table 1) despite minor stochastic variations in theirintrinsicparameters.Thisdeterminationofgenecircuitdesigncriteriain the
44、 present context of a fast, robust and tunable oscillator sets thestageforthedesignofapplicationssuchasexpressionschemesthatarecapable of circumventing cellular adaptability, centralized clocks thatcoordinate intracellular behaviour, and reverse-engineering plat-forms21that measure the global respon
45、se of the genome to an oscil-latory perturbation.METHODS SUMMARYThe dual-feedback oscillator circuit was constructed by placing araC, lacI andyemGFP under the control of the hybrid plac/ara-1promoter14in three separatetranscriptional cassettes. An ssrA degradation tag22was added to each gene todecre
46、ase protein lifetime and to increase temporal resolution. These transcrip-tionalcassetteswereplacedontwomodularplasmids14andco-transformedintoan DaraCDlacI E. coli strain. The negative feedback oscillator circuit was con-structedbyplacingssrA-taggedlacIandyemGFPunderthecontrolofthepLlacO-1promoter14
47、in two separate transcriptional cassettes, which were incorporatedonto two modular plasmids and co-transformed into a DlacI strain. Cells wereab0 120 180Time (min) IPTGpLlacO-1pLlacO-160lacIyemGFPFigure 3 | Anoscillatorwithnopositivefeedbackloop. a,Networkdiagramof the negative feedback oscillator.
48、This oscillator is similar to the dual-feedback oscillator except that the hybrid promoter pLlacO-1(ref. 14) givesexpression of lacI and yemGFP in the absence of LacI or in the presence ofIPTG without requiring an activator. b, Single-cell density map trajectoriesfor cells containing this oscillator
49、 (see Supplementary Movie 11 andSupplementary Fig. 5).Box 1 | Dynamic modelling of the dual-feedback oscillator circuitWe used standard techniques to construct both stochastic anddeterministic computational models3,2528based on the sameunderlying biochemical reactions illustrated in Fig. 4a (seeSupplementary Information for full details of modelling). Although theinteraction between transcription factors and the DNA is generallyquitecomplicatedtomodelindetail29,weusedexperimentalinductioncurves to calibrate the induction levels in the reactions describing thenetwork (Supplem
