a sensing array of radically coupled genetic 'biopixels' · 2017-08-22 · a sensing...

ARTICLEdoi:10.1038/nature10722

A sensing array of radically coupledgenetic ‘biopixels’Arthur Prindle1*, Phillip Samayoa2*, Ivan Razinkov1, Tal Danino1, Lev S. Tsimring3 & Jeff Hasty1,2,3,4

Although there has been considerable progress in the development of engineering principles for synthetic biology, asubstantial challenge is the construction of robust circuits in a noisy cellular environment. Such an environment leads toconsiderable intercellular variability in circuit behaviour, which can hinder functionality at the colony level. Here weengineer the synchronization of thousands of oscillating colony ‘biopixels’ over centimetre-length scales through theuse of synergistic intercellular coupling involving quorum sensing within a colony and gas-phase redox signallingbetween colonies. We use this platform to construct a liquid crystal display (LCD)-like macroscopic clock that can beused to sense arsenic via modulation of the oscillatory period. Given the repertoire of sensing capabilities of bacteria suchas Escherichia coli, the ability to coordinate their behaviour over large length scales sets the stage for the construction oflow cost genetic biosensors that are capable of detecting heavy metals and pathogens in the field.

Synthetic biology can be broadly broken down into the ‘top-down’synthesis of genomes1 and the ‘bottom-up’ engineering of relativelysmall genetic circuits2–10. In the field of genetic circuits, toggleswitches11 and oscillators12 have progressed into triggers13, counters14

and synchronized clocks15. Sensors have arisen as a major focus in thecontext of biotechnology6,16,17, while oscillators have provided insightsinto the basic-science functionality of cyclic regulatory processes18–20.A common theme is the concurrent development of mathematicalmodelling that can be used for experimental design and characteriza-tion, as in physics and the engineering disciplines.

The synchronization of genetic clocks provides a particularlyattractive avenue for synthetic biology applications. Oscillationspermeate science and technology in a number of disciplines, withfamiliar examples including alternating current (AC) power21, theglobal positioning system (GPS)22 and lasers23. These technologieshave demonstrated that operating in the frequency domain can offerconsiderable advantages over steady-state designs in terms ofinformation gathering and transmission. In particular, oscillatorysensors confer a number of advantages to traditional ones24, asfrequency is easily digitized and can be quickly updated with repeatedmeasurements. For sensors that use optical reporters, measurementsof frequency are less sensitive to experimental factors such as beampower and exposure time than intensity measurements, which mustbe normalized and calibrated.

Although the bottom-up approach to synthetic biology is increas-ingly benefiting from DNA synthesis technologies, the general designprinciples are still evolving. In this context, a substantial challenge isthe construction of robust circuits in a cellular environment that isgoverned by noisy processes such as random bursts of transcriptionand translation25–29. Such an environment leads to considerable inter-cellular variability in circuit behaviour, which can impede coherentfunctionality at the colony level. An ideal design strategy for reducingvariability across a cellular population would involve both strong andlong-range coupling that would instantaneously synchronize the res-ponse of millions of cells. Quorum sensing typically involves strongintercellular coupling over tens of micrometres8,15,30, yet the relatively

slow diffusion time of molecular communication through cellularmedia leads to signalling delays over millimetre scales. Faster com-munication mechanisms, such as those mediated in the gas phase,may increase the length scale for instantaneous communication, butare comparatively weak and short lived because the vapour speciesmore readily disperse.

Synergistic synchronizationTo develop a frequency-modulated biosensor, we designed a genenetwork capable of synchronizing genetic oscillations across multiplescales (Fig. 1a and Supplementary Fig. 1). We constructed an LCD-like microfluidic31 array that allows many separate colonies of sensingbacteria to grow and communicate rapidly by gas exchange (Fig. 1b, cand Supplementary Fig. 9). As previous work15 has demonstrated thatcoupling through quorum sensing leads to incoherent oscillations atthe millimetre scale, this mode of cellular communication is tooslow for the generation of macroscopic synchronized oscillations.However, the slower quorum sensing can be used to synchronize smalllocal colonies, provided there is a second level of design that involvesfaster communication for coordination between the colonies.Therefore, rather than attempting to engineer a sensor from a singlelarge-colony oscillator, we wired together thousands of small oscil-lating colonies, or ‘biopixels’, in a microfluidic array. Coupling betweenbiopixels involves redox signalling by hydrogen peroxide (H2O2) andthe native redox sensing machineries of E. coli. The two couplingmechanisms act synergistically in the sense that the stronger, yetshort-range, quorum sensing is necessary to coherently synchronizethe weaker, yet long-range, redox signalling. Using this method wedemonstrate synchronization of approximately 2.5 million cells acrossa distance of 5 mm, over 1,000 times the length of an individual cell(Fig. 1c, d and Supplementary Movies 1 and 2). This degree of syn-chronization yields extremely consistent oscillations, with a temporalaccuracy of about 2 min compared to 5–10 min for a single oscillator15

(Fig. 1d).The global synchronization mechanism is comprised of two modes

of communication that work on different scales. The quorum-sensing

*These authors contributed equally to this work.

1Department of Bioengineering, University of California, San Diego, La Jolla, California 92093, USA. 2Bioinformatics Program, University of California, San Diego, La Jolla, California 92093, USA. 3BioCircuitsInstitute, University of California, San Diego, La Jolla, California 92093, USA. 4Molecular Biology Section, Division of Biological Science, University of California, San Diego, La Jolla, California 92093, USA.

0 0 M O N T H 2 0 1 1 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2011

www.nature.com/doifinder/10.1038/nature10722

machinery (LuxI, AiiA) uses an acyl-homoserine lactone (AHL) tomediate intracolony synchronization. In our device, the degree towhich neighbouring colonies are able to influence each other viaAHL diffusion is negligible owing to the high media channel flowrates. Instead, we engineered the cells to communicate via gasexchange by placing a copy of the gene coding for NADH dehydro-genase II (ndh) under the control of an additional lux promoter.NDH-2 is a membrane-bound respiratory enzyme that produceslow levels of H2O2 and superoxide (O2

2)32. As H2O2 vapour is ableto pass through the 25-mm oxygen-permeable polydimethylsiloxane(PDMS) walls that separate adjacent colonies, periodic production ofNDH-2 yields periodic exchange of H2O2 between biopixels. WhenH2O2 enters the cell, it transiently changes its redox state, interactingwith our synthetic circuit through the native aerobic response controlsystems, including ArcAB, which has a binding site in the lux pro-moter region33,34. Under normal conditions, ArcAB is partially activeso lux is partially repressed. In contrast, oxidizing conditions triggeredby H2O2 inactivate ArcAB, relieving this repression. Each oscillatoryburst promotes firing in neighbouring colonies by relieving repressionon the lux promoter. This constitutes an additional positive feedbackthat rapidly synchronizes the population (Supplementary Fig. 2 andSupplementary Movie 1).

We investigated the effects of catalase and superoxide dismutase(SOD) to probe the nature of H2O2 communication. When a popu-lation of synchronized colonies was exposed to a step increase of200 U ml21 catalase, an enzyme that rapidly degrades extracellularH2O2

35, synchronization was broken and colonies continued to oscil-late individually (Supplementary Fig. 3). As the cell membrane isimpermeable to catalase, asynchronous colony oscillations confirmthat communication between colonies depends on external H2O2

whereas oscillations within a colony do not. Conversely, when weenhanced the rate of superoxide conversion to H2O2 by expressing

sodA36,37 from an additional lux promoter, colonies quickly fired in aspatial wave and failed to oscillate further despite no changes to growthrate or cell viability (Supplementary Fig. 4). Because H2O2 is producedinternal to the cell, this confirms that H2O2 is capable of escaping thecell and activating lux-regulated genes in neighbouring colonies viadiffusion. The apparent higher output of H2O2 by SOD as compared toNDH-2 is probably due to its very high catalytic efficiency38. Lastly, weobserved synchronization between arrays of traps even when theywere fluidically isolated but held in close proximity (SupplementaryFig. 5). These devices share no common fluid sources or channels,making communication by dissolved molecules like AHL impossible.Taken together, these results confirm that gaseous H2O2 is the mode ofcommunication between oscillating colonies.

On the basis of our understanding of the mechanism for globalsynchronization, we expected that we could simplify the circuitry byeliminating ndh and achieve the same effect with intermittent bursts ofhigh-intensity blue light. In this design, the GFP molecule acts as aphotosensitizer, releasing free radicals upon exposure that producereactive oxygen species (ROS) including H2O2

39. At the peak of oscil-lation, considerable vapour-phase H2O2 is produced by exposing GFP-containing cells to fluorescent light. Conversely, at the trough of oscil-lation, cells contain almost no GFP, and therefore produce very littleH2O2 upon fluorescing. Bursts of light thus generate bursts of H2O2

vapour whose concentration depends on the oscillating GFP level, justas periodic production of NDH-2 did previously. Indeed, this strategywas similarly able to synchronize our sensor array (Fig. 1d andSupplementary Movie 2). Numerous controls were performed toensure that synchronized oscillations did not occur at low fluorescenceintensities (Supplementary Fig. 6 and Supplementary Movie 9).

To probe this mode of synchronization, we investigated the effects ofthiourea and the antibiotics ampicillin and kanamycin. When a syn-chronized population of colonies was exposed to 35 mM thiourea, a

AHL

LuxR-AHL

luxI aiiA

ndh sfGFP

H2O2

Reporter

Oscillator

Coupling

5,000 cells per biopixel

2.5 million total cells

00

0.5

1

100 200 300 400 500 600 700 800 900 1,000

Time (min)

Biopixel GFP

Mean

00

250

500

100 200 300 400 500 600 700 800 900 1,000

Time (min)F

luo

rescence

(arb

itra

ry u

nits)

Bio

pix

el n

um

ber

a c

b d

Dissolved AHL

H2O2 vapour

Media

Figure 1 | Sensing array of radically coupled genetic biopixels. a, Networkdiagram. The luxI promoter drives expression of luxI, aiiA, ndh and sfGFP(superfolder variant of GFP) in four identical transcription modules. Thequorum-sensing genes luxI and aiiA generate synchronized oscillations withina colony via AHL. The ndh gene codes for NDH-2, an enzyme that generatesH2O2 vapour, which is an additional activator of the luxI promoter. H2O2 iscapable of migrating between colonies and synchronizing them. b, Conceptual

design of the sensing array. AHL diffuses within colonies while H2O2 migratesbetween adjacent colonies through the PDMS. Arsenite-containing media ispassed in through the parallel feeding channels. c, Fluorescent image of an arrayof 500 E. coli biopixels containing about 2.5 million cells. Inset, bright-field andfluorescent images display a biopixel of 5,000 cells. d, Heat map and trajectoriesdepicting time-lapse output of 500 individual biopixels undergoing rapidsynchronization. Sampling time is 2 min.

RESEARCH ARTICLE

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 1


potent radical quencher40,41, we observed sharply decaying synchronizedoscillations whereas growth rate and cell viability were unaffected(Supplementary Fig. 7). This suggests that without radical species,oscillations cannot be produced. Next, we ran a series of experimentsswitching the antibiotic resistance genes on our plasmids. We notedthat radical-producing antibiotics42, particularly ampicillin, signifi-cantly reduced the degree of synchronization, showing that an excessof radical species also hinders communication (Supplementary Fig. 8).As our final constructs included a plasmid with kanamycin resistance,which was also found to produce some radicals, we used full (50mgml21) selection when growing up the cells but very low (5mg ml21)selection during the experimental run. Persistence of oscillations,sequencing, and subsequent growth in full selection following therun confirmed the presence of all three plasmids despite this lowexperimental selection. Catalase and SOD results were identical tothose with NDH-2 synchronization (not shown). These results showthat fluorescence-mediated synchronization involves the productionof radical species after fluorescence exposure and communication viaH2O2.

Sensing array of biopixelsWith a platform for generating consistent and readily detectable oscil-lations, we sought to use the circuit to engineer an arsenic-sensingmacroscopic biosensor. We rewired the network to include an extracopy of the positive-feedback element, the AHL synthase LuxI, underthe control of a native arsenite-responsive promoter that is repressedby ArsR in the absence of arsenite (Fig. 2a, right). When arsenite is notpresent in the media, supplementary luxI is not transcribed and thecircuit functions normally, generating baseline oscillations. However,the addition of trace amounts of arsenite relieves this repression and

allows supplementary luxI to be transcribed, increasing the oscillatoryamplitude and period. Tuning the level of LuxI by varying arseniteconcentration results in clear changes to the oscillatory period (Fig. 2band Supplementary Movie 2). To determine the range of detection, weswept arsenite concentrations from 0–1 mM and measured the oscil-latory period (Fig. 2c, top). Using statistical methods (SupplementaryMethods), we generated a sensor calibration curve (Fig. 2c, bottom)that depicts the maximum possible arsenite concentration present(a 5 95%) for a given measured period. This curve is an illustrationof how data generated by our array would be used to measure arseniteconcentrations in an unknown sample using our device. Our systemwas able to reliably quantify arsenite levels as low as 0.2 mM, belowthe 0.5 mM World Health Organization-recommended level fordeveloping nations43.

As an alternative sensing strategy, we rewired the network toinclude a copy of the luxR gene controlled by an arsenite-responsivepromoter while removing it from the rest of the circuit (Fig. 2a, left).Because the LuxR–AHL complex must be present to activate the luxpromoter30, cells produce no LuxR when the media is free of arsenite,generating no fluorescence or oscillations. The addition of arsenitestimulates the production of LuxR, restoring circuit function andproducing clear, synchronized oscillations (Fig. 2d and Supplemen-tary Movie 3). This ON/OFF detection system has a threshold of0.25 mM, a detection limit that can be adjusted by changing the copynumber, ribosome binding site (RBS) strength, or promoter strengthof the sensing plasmid (Supplementary Methods).

The sensing array is also capable of producing complex behavioursarising from the dynamic interaction of cellular colonies. By makingmodifications to the size, number and arrangement of biopixels in thedevice, we are able to markedly alter the output waveforms. For

Arsenite

1 2

luxR

ArsR

luxI

ArsR

luxR

0 400 800 1,200Time (min)

Mean fl

uo

rescence (arb

itra

ry u

nits)

– Arsenite + 0.8 μM Arsenite

0 200 400 600 800

0.2

0.4

0.6

0.8

1

Time (min)

Flu

ore

scence (arb

itra

ry u

nits)

Thresholding Period modulation

a b

d

c

0 0.2 0.4 0.6 0.8 1

68

72

76

80

84

Arsenite (μM)

Perio

d (m

in)

70 72 74 76 780

0.2

0.4

0.6

0.8

1

Period (min)

Ars

enite (μM

)

0.2

0.4

0.6

0.8

1

– Arsenite + 0.25 μM Arsenite

Figure 2 | Frequency-modulated genetic biosensor. a, Network diagramsdepicting two constructed sensing modules. In thresholding (1), the luxR geneis removed from the oscillator network and supplemented by a new copy drivenby an arsenite-responsive promoter. In period modulation (2), a supplementaryluxI gene tagged for increased degradation is driven by the arsenic-responsivepromoter, which affects the period of oscillation. b, A sample periodmodulation sensor output following a step increase of 0.8mM arsenite.Oscillatory period increases from 69 min to 79 min. c, Top, period versus

arsenite concentration for the sensor array. Error bars indicate 6 1 standarddeviation averaged over 500 biopixel trajectories. Dotted line represents model-predicted curve. Bottom, sensor calibration curve generated from experimentaldata. Points indicate the maximum arsenite level with 95% certainty for a givenmeasured period as determined statistically from experimental data.d, Thresholder output following a step increase of 0.25 mM arsenite. A markedshift from rest to oscillatory behaviour is observed within 20 min after theaddition of arsenite.

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 1 | V O L 0 0 0 | N A T U R E | 3


example, when we constructed a device in which trap separation dis-tance is increased (45mm versus 25mm), we observed local anti-phasesynchronization between neighbouring colonies (Fig. 3d, top right). Toexplore this phenomenon on a larger scale, we constructed a device thatcontains an array of 416 traps constructed according to the specifica-tions above. In these experiments, we observe initial global synchron-ization that gradually falls into local anti-phase synchronization acrossthe array (Fig. 3d, middle. and Supplementary Movie 4). Phase align-ment is maintained over at least 48 h, with patches of synchronizationtypically 3–6 colonies in size. Alternatively, by changing dimensionssuch that the array contains traps of two slightly different sizes, weobserve a 1:2 resonance synchronization where larger traps pulse atdouble the frequency of smaller traps while maintaining synchroniza-tion (Fig. 3d, top left and Supplementary Movie 7). Lastly, when LuxRis limited, as in the thresholding scheme, we observe synchronizedoscillations of alternating large and small peaks in both experimentand model (Supplementary Fig. 12). Our computational model (Box 1)captures these effects (Fig. 3d, bottom, and Supplementary Figs 11and 12) and indicates that further array manipulation will yield new,richer dynamics that could not be produced directly by changing cir-cuit structure.

Although our sensor array is capable of performing a variety ofcomplex functions in the laboratory, adapting this technology to areal-world device will require the elimination of the expensive andbulky microscopy equipment. However, measuring genetic oscilla-tions in the absence of any magnification or powerful illuminationwill require an even further increased signal. Using this mechanism ofglobal synchronization, we were able to scale up to a 24 mm 3 12 mmarray that houses over 12,000 communicating biopixels (Fig. 4a).Synchronization is maintained across the entire array, a distance over5,000 times the length of an individual cell, using an inexpensive light-emitting diode (LED; Fig. 4b, c and Supplementary Movie 5). The

signal strength generated by the large number of cells in the array(about 50 million) will allow us to adapt the device to function as ahandheld sensor. In our conceptual design (Fig. 4d), the sensor will

a

b c

d

Time (min)

Bio

pix

el num

ber

2000

200

400

400 600 800 1,000

0 200 400 600 800

Time (min)

Arsenic

GFP

200 400 600

40

80

120

160

200

Flu

ore

scence

(arb

itra

ry u

nits)

Flu

ore

scence

(arb

itra

ry u

nits)

Flu

ore

scence

(arb

itra

ry u

nits)

Time (min)

0

Low H2O2 (catalase) Moderate H2O2 High H2O2 (SOD)

600 800 1,0000

100

200

300

400

Time (min)

200 400 600

40

80

120

160

200

Time (min)

0 200 400 600

40

80

120

160

200

Time (min)

0

0.2

0.4

0.6

0.8

1

0 200 400 600 800

Time (min)

0.2

0.4

0.6

0.8

1

600 800 1,0000

100

200

300

400

Time (min)

Flu

ore

scence

(arb

itra

ry u

nits)

Flu

ore

scence

(arb

itra

ry u

nits)

0

0.5

1

0

0.5

1

Time (min)600 800 1,000

Time (min)600 800 1,000

Figure 3 | Computational modelling of radical synchronization andbiosensing. a, Time series of a population of biopixels producing varyingamounts of H2O2 vapour. Synchronization occurs only for moderate levelswhereas high levels lock ON and low levels oscillate asynchronously. b, Atypical time series for our period modulation sensor undergoing a step increaseof arsenite. Oscillations increase in both amplitude and period. c, A typical timeseries output for the thresholding sensor. Oscillations arise after the addition ofarsenite. d, Experimental and computational output depicting complex

dynamic behaviours between neighbouring traps. Top, 1:2 resonance and anti-phase synchronization observed when trap size (left, black/blue 5 95mm depthand red/magenta 5 85mm depth) and separation distance (right, same colours)are modified experimentally. Middle, scaled-up array experimental data forincreased trap separation experiments demonstrating anti-phasesynchronization. Bottom, computational model trajectories depicting 1:2resonance and anti-phase synchronization when trap size (same colours asexperimental data) and separation distance are changed.

Time (min)

Bio

pix

el num

ber

120 240 360

2,000

4,000

6,000

8,000

10,000

05 cm

5 mm

t = 0 min

t = 30 min

t = 60 min

t = 90 min

t = 120 min

t = 150 min

a

d

b c

12,000

5

2

4

1

3

Figure 4 | Radical synchronization on a macroscopic scale. a, The scaled-uparray is 24 mm 3 12 mm and houses over 12,000 biopixels that containapproximately 50 million total cells when filled. b, Global synchronization ismaintained across the array. Heat map of individual trajectories of all 12,224oscillating biopixels. c, Image series depicting global synchronization andoscillation for the macroscopic array. Each image is produced by stitching 72fields of view imaged at 34 magnification. d, Schematic diagram illustratingour design for a handheld device using the sensing array. An LED (1) excites thearray (2) and emitted light is collected by a photodetector (3), analysed by anonboard processor (4), and displayed graphically (5).

RESEARCH ARTICLE

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 1


continuously read the oscillatory frequency using off-the-shelf elec-tronic components costing less than 50 dollars.

There have been many examples of bacteria-based biosensors44–46,usually involving an optical reporter driven by a toxin-responsive pro-moter. Because optical intensity readings are sensitive to imaging con-ditions like beam power and exposure time, measurements musttypically be normalized and calibrated. Measuring the period of oscil-lation allows us to avoid these issues because peak-to-peak time doesnot depend on individual peak intensity. Also, oscillations produced atthe colony level effectively decouple the signal from the growth state ofindividual cells, which can also affect fluorescence intensity. By using adynamic readout that depends on communication between biopixels,we scan and tune potential output signals by changing device para-meters rather than redesigning the underlying circuit. For example, wemight design a new sensing scheme in which oscillations synchronizewith the addition of some toxin and shift to anti-phase or resonantsynchronization when critical toxin levels are present.

Scaling up synthetic biologyBy nesting two modes of communication we are able to expand thescale over which individual cells are coordinated and increase thecomplexity of their interaction. Indeed, there are many familiarexamples of hierarchical systems. Airline routes are often designedsuch that small airports are connected locally to larger hubs that areconnected internationally. It would neither be feasible nor desirable toconnect every airport together. Similarly, individual cells communicate

locally by one method, generating impulses large enough to enablecolonies to communicate globally by another. Nesting communicationmechanisms in this way may allow us to better scale up syntheticcircuits of different types, such as switches and logic gates, pavingthe way for the next generation of synthetic biology pursuits.

METHODS SUMMARYStrains and plasmids. The plasmids were constructed using a PCR-based cloningstrategy48 in which the origin of replication, antibiotic resistance, and circuit geneswere assembled in different combinations. The ndh and sodA genes were amp-lified directly from the native E. coli genome by PCR. Various arsenite-responsivepromoters were tested, including a recently reported synthetic version49, but thefinal design uses the native E. coli version. Promoter output was tuned bychanging the RBS sequence and quantified using flow cytometry. All circuitcomponents except luxR were tagged by PCR with a carboxy-terminal ssrA tag(AANDENYALAA)50 for fast degradation.Microfluidics and microscopy. Image acquisition was performed on a NikonEclipse TI epifluorescent inverted microscope outfitted with fluorescence filtercubes optimized for GFP imaging and a phase-contrast-based autofocus algorithm.Images were acquired using an Andor Clara cooled CCD camera or Andor DU-897EMCCD camera, both controlled by Nikon Elements software. Images wereacquired every 2 min in phase contrast and fluorescence. The cells were imagedinside a microfluidic device with an upstream switch, with the ability to mix orswitch between two different media sources. A custom application written inLabVIEW (National Instruments) controlled linear actuators, to which tworeservoirs of arsenite-containing and pure medium were attached. Using thisalgorithm, arsenite concentration was dynamically varied to probe sensor output.

The microfluidic experiments were performed as previously described15.Briefly, 50ml of an overnight culture was diluted in 50 ml of LB medium(Difco) plus antibiotics the day of the experiment. When cells reached an opticaldensity (OD600 nm) of 0.1, cells were spun down and resuspended in 5 ml of freshmedia and loaded into the device.

Received 2 September; accepted 15 November 2011.

Published online 18 December 2011.

1. Gibson, D. et al. Creation of a bacterial cell controlled by a chemically synthesizedgenome. Science 329, 52–56 (2010).

2. Hasty, J., McMillen, D. & Collins, J. J. Engineered gene circuits. Nature 420,224–230 (2002).

3. Sprinzak, D. & Elowitz, M. B. Reconstruction of genetic circuits. Nature 438,443–448 (2005).

4. Endy, D. Foundations for engineering biology. Nature 438, 449–453 (2005).5. Ellis, T., Wang, X. & Collins, J. Diversity-based, model-guided construction of

synthetic gene networks with predicted functions. Nature Biotechnol. 27, 465–471(2009).

6. Kobayashi, H. et al. Programmable cells: interfacing natural and engineered genenetworks. Proc. Natl Acad. Sci. USA 101, 8414–8419 (2004).

7. You, L., Cox, R. S. III, Weiss, R. & Arnold, F. H. Programmed population control bycell–cell communication and regulated killing. Nature 428, 868–871 (2004).

8. Basu, S., Gerchman, Y., Collins, C. H., Arnold, F. H. & Weiss, R. A syntheticmulticellular system for programmed pattern formation. Nature 434, 1130–1134(2005).

9. Mukherji, S. & Van Oudenaarden, A. Synthetic biology: understanding biologicaldesign from synthetic circuits. Nature Rev. Genet. 10, 859–871 (2009).

10. Grilly, C., Stricker, J., Pang, W., Bennett, M. & Hasty, J. A synthetic gene network fortuning protein degradation in Saccharomyces cerevisiae. Mol. Syst. Biol. 3, 127(2007).

11. Gardner, T. S., Cantor, C. R. & Collins, J. J. Construction of a genetic toggle switch inEscherichia coli. Nature 403, 339–342 (2000).

12. Elowitz, M. B. & Leibler, S. A synthetic oscillatory network of transcriptionalregulators. Nature 403, 335–338 (2000).

13. Lu, T. & Collins, J. Dispersing biofilms with engineered enzymatic bacteriophage.Proc. Natl Acad. Sci. USA 104, 11197–11202 (2007).

14. Friedland, A. et al. Synthetic gene networks that count. Science 324, 1199–1202(2009).

15. Danino, T., Mondragon-Palomino, O., Tsimring, L. & Hasty, J. A synchronizedquorum of genetic clocks. Nature 463, 326–330 (2010).

16. Tamsir, A., Tabor, J. & Voigt, C. Robust multicellular computing using geneticallyencoded NOR gates and chemical ‘wires’. Nature 469, 212–215 (2011).

17. Tabor, J. et al. A synthetic genetic edge detection program. Cell 137, 1272–1281(2009).

18. Stricker, J. et al. A fast, robust and tunable synthetic gene oscillator. Nature 456,516–519 (2008).

19. Mondragon-Palomino, O., Danino, T., Selimkhanov, J., Tsimring, L. & Hasty, J.Entrainment of a population of synthetic genetic oscillators. Science 333,1315–1319 (2011).

20. Tigges, M., Marquez-Lago, T., Stelling, J. & Fussenegger, M. A tunable syntheticmammalian oscillator. Nature 457, 309–312 (2009).

BOX 1

Computational modellingOur model of the frequency-modulated biosensor is based on apreviously described model for the quorum-sensing synchronizedoscillator15. In addition to the reactions reflected in that model, weinclude the arsenite-induced production and degradation of LuxI and/or LuxR. From the biochemical reactions, we derived a set of delay-differential equations to be used as our model. These delayedreactions mimic the complex cascade of processes (transcription,translation, maturation, etc.) leading to formation of functionalproteins. As expected, our model predicts oscillations that changefrequency when changes in arsenite occur (Figs 2c and 3b). Theamplitude and period of the oscillations both depend on theconcentrations of the toxin. We then modified the model to describethe LuxR-based detection system. Our model predicts a markedtransition from rest to oscillationsuponaddition of arsenite, consistentwith experimental observations (Fig. 3c).

The multi-scale nature of communication in our array allows us totreat colony and array-level dynamics separately; in the latter, arseniteaffects the quorum-sensing machinery of a colony, producingchanges to oscillatory period that propagate between biopixels in thearray. To describe quantitatively the mechanisms drivingsynchronization at the array level, we treat each colony as a singleoscillator that acts according to degrade-and-fire kinetics47. We alsoinclude the production of H2O2 and its interaction with neighbouringcolonies by two-dimensional diffusion. Using this model we identifiedthree regimes that correlate well with experimental observations(Fig. 3a). When the effectiveproduction of H2O2 is low, as with catalase,we observe unsynchronized oscillations owing to constant, mildrepressionof the luxpromoter viaArcAB (Fig. 3a, left). In contrast,whenH2O2 production is very high, neighbouring colonies rapidly fire insuccession and remain on because of the permanent activation of thelux promoter, consistent with the SOD experiment (Fig. 3a, right).Finally, at intermediate H2O2, we observe globally synchronizedoscillations (Fig. 3a, middle). As colonies are moved further apart,synchronicity breaks owing to slowed migration of H2O2

(Supplementary Fig. 10).

ARTICLE RESEARCH

0 0 M O N T H 2 0 1 1 | V O L 0 0 0 | N A T U R E | 5


21. Westinghouse, G. System of electrical distribution. US patent 373 035 (1887).22. Lewandowski, W., Azoubib, J. & Klepczynski, W.GPS: primary tool for time transfer.

Proc. IEEE 87, 163–172 (1999).23. Vladimirov, A., Kozyreff, G. & Mandel, P. Synchronization of weakly stable

oscillators and semiconductor laser arrays. Europhys. Lett. 61, 613 (2003).24. Gast, T. Sensors with oscillating elements. J. Phys. E: Sci. Instrum. 18, 783 (1985).25. Ozbudak, E. M., Thattai, M., Kurtser, I., Grossman, A. D. & van Oudenaarden, A.

Regulation of noise in the expression of a single gene. Nature Genet. 31, 69–73(2002).

26. Elowitz, M. B., Levine, A. J., Siggia, E. D. & Swain, P. S. Stochastic gene expression ina single cell. Science 297, 1183–1186 (2002).

27. Golding, I., Paulsson, J., Zawilski, S. & Cox, E. Real-time kinetics of gene activity inindividual bacteria. Cell 123, 1025–1036 (2005).

28. Blake, W. et al. Phenotypic consequences of promoter-mediated transcriptionalnoise. Mol. Cell 24, 853–865 (2006).

29. Austin, D. et al. Gene network shaping of inherent noise spectra. Nature 439,608–611 (2006).

30. Waters, C. & Bassler, B. Quorum sensing: cell-to-cell communication in bacteria.Annu. Rev. Cell Dev. Biol. 21, 319–346 (2005).

31. Ferry, M., Razinkov, I. & Hasty, J. Microfluidics for synthetic biology from design toexecution. Methods Enzymol. 497, 295 (2011).

32. Messner, K. & Imlay, J. The identification of primary sites of superoxide andhydrogen peroxide formation in the aerobic respiratory chain and sulfitereductase complex of Escherichia coli. J. Biol. Chem. 274, 10119–10128 (1999).

33. Bose, J. L. et al. Bioluminescence in Vibrio fischeri is controlled by the redox-responsive regulator arca. Mol. Microbiol. 65, 538–553 (2007).

34. Georgellis, D., Kwon, O. & Lin, E. Quinones as the redox signal for the arc two-component system of bacteria. Science 292, 2314–2316 (2001).

35. Seaver, L. & Imlay, J. Hydrogen peroxide fluxes and compartmentalization insidegrowing Escherichia coli. J. Bacteriol. 183, 7182–7189 (2001).

36. Fridovich, I. The biology of oxygen radicals. Science 201, 875–880 (1978).37. McCord, J. & Fridovich, I. Superoxide dismutase. J. Biol. Chem. 244, 6049–6055

(1969).38. Berg, J., Tymoczko, J. L. & Stryer, L. Biochemistry (W.H. Freeman, 2006).39. Remington, S. Fluorescent proteins: maturation, photochemistry and

photophysics. Curr. Opin. Struct. Biol. 16, 714–721 (2006).40. Kelner, M., Bagnell, R. & Welch, K. Thioureas react with superoxide radicals to yield

a sulfhydryl compound. explanation for protective effect against paraquat. J. Biol.Chem. 265, 1306–1311 (1990).

41. Touati, D., Jacques, M., Tardat, B., Bouchard, L. & Despied, S. Lethal oxidativedamage and mutagenesis are generated by iron in delta fur mutants of

Escherichia coli: protective role of superoxide dismutase. J. Bacteriol. 177,2305–2314 (1995).

42. Kohanski, M. A., DePristo, M. A. & Collins, J. J. Sublethal antibiotic treatment leadsto multidrug resistance via radical-induced mutagenesis. Mol. Cell 37, 311–320(2010).

43. Nordstrom, D. Worldwide occurrences of arsenic in ground water. Science 296,2143 (2002).

44. van der Meer, J. & Belkin, S. Where microbiology meets microengineering: designand applications of reporter bacteria. Nature Rev. Microbiol. 8, 511–522 (2010).

45. Daunert, S. et al. Genetically engineered whole-cell sensing systems: couplingbiological recognition with reporter genes. Chem. Rev. 100, 2705–2738 (2000).

46. Leveau, J. & Lindow, S. Bioreporters in microbial ecology. Curr. Opin. Microbiol. 5,259–265 (2002).

47. Mather, W., Bennett, M., Hasty, J. & Tsimring, L. Delay-induced degrade-and-fireoscillations in small genetic circuits. Phys. Rev. Lett. 102, 068105 (2009).

48. Quan, J. & Tian, J. Circular polymerase extension cloning of complex gene librariesand pathways. PLoS ONE 4, e6441 (2009).

49. Stocker, J.et al.Developmentof a set of simplebacterial biosensors forquantitativeand rapid measurements of arsenite and arsenate in potable water. Environ. Sci.Technol. 37, 4743–4750 (2003).

50. Keiler, K., Waller, P. & Sauer, R. Role of a peptide tagging system in degradation ofproteins synthesized from damaged messenger RNA. Science 271, 990–993(1996).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements This work was supported by the National Institutes of Health andGeneral Medicine (R01GM69811), the San Diego Center for Systems Biology(P50GM085764), the DoD NDSEG (A.P.) and NSF Graduate Research (P.S.) FellowshipPrograms. L.T. was supported, in part, by the Office of Naval Research (MURIN00014-07-0741). We would like to thank J. Imlay, S. Ali and J. Collins for helpfuldiscussions and J. Hickman, B. Taylor and K. Lomax for their help with illustrations.

Author Contributions All authors contributed extensively to the work presented in thispaper. A.P. and P.S. are equally contributing first authors.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of this article atwww.nature.com/nature. Correspondence and requests for materials should beaddressed to J.H. ([email protected]).

RESEARCH ARTICLE

6 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 1


www.nature.com/nature

www.nature.com/reprints

www.nature.com/nature

mailto:[email protected]

W W W. N A T U R E . C O M / N A T U R E | 1

SUPPLEMENTARY INFORMATIONdoi:10.1038/nature10722

Supplementary InformationSensing array of radically coupled genetic biopixels

Arthur Prindle,1� Phillip Samayoa,2� Ivan Razinkov,1 Tal Danino,1 Lev S. Tsimring,3

and Jeff Hasty1,2,3,4,5

November 8, 2011

1 Department of Bioengineering, University of California, San Diego, San Diego, California, USA.2 Bioinformatics Program, University of California, San Diego, La Jolla, California, USA.3 BioCircuits Institute, University of California, San Diego, San Diego, California, USA.4 Molecular Biology Section, Division of Biological Science, University of California, San Diego, La Jolla,CA 92093, USA5 Corresponding Author. Molecular Biology Section, Division of Biological Science, University of Califor-nia, San Diego, Mailcode 0412, La Jolla, CA 92093-0412, USA. Telephone: 858 822 3442. Fax: 858 5345722. Email: [email protected]

� These authors contributed equally to this work.

Plasmid Construction

The oscillator plasmids were constructed by modifying the constructs used in a previous study(1). The antibiotic resistance genes of pTD103AiiA was switched to chloramphenicol. The re-porter protein on pTD103LuxI/GFP was switched to a recently reported superfolding greenfluorescent protein, sfGFP (2). The ndh and sodA genes were amplified directly from the nativeE. coli genome by PCR. Promoter output was tuned by changing the RBS sequence and quantifiedusing flow cytometry. We initially constructed the sensing plasmid with a published syntheticbackground-reduced version that contains additional ArsR operator sites(3) but failed to produceenough LuxR. To increase LuxR output, we reverted to the native promoter sequence, switchedthe RBS to that of pZ plasmids(? ), and increased the copy number by a factor of 5 by switch-ing to a mutated SC101 origin of replication. All circuit components except LuxR were taggedby PCR with a carboxy-terminal ssrA tag (AANDENYALAA) (4) for fast degradation. Modu-lar pieces (resistance genes, promoters, origins, and ORFs) were assembled using a PCR-basedcloning scheme named CPEC (5).

1��

��

aiiA LAA

Cm

luxIp

p15A

T1

T0

��

��

��

��

��

��

��

luxR

Spect

pArsR

Lutz RBS

pSC101m

T1

T0

��

��

��

��

��

��

sfGFP LAA

LuxI LAA

Kan

luxIp

luxIp

colE1

T1

T1

T0

��

��

��

��

LuxI-LAA

Spect

pArsR

Lutz RBS

pSC101

T1

T0

��

��

��

��

��

��

LuxR

LuxR

sfGFP

LuxI

Kan

pLuxI

pLuxI

colE1

T1

T1

T0

��

��

��

��

��

luxR

aiiA LAA

Cm

luxIp

p15A

T1

T0

��

��

��

��

��

��

LuxR

ndhII LAA

Spect

LuxIp

pSC101m

T0

T1

��

��

��

��

��

��

LuxR

sodA LAA

Spect

LuxIp

pSC101m

T1

T0

��

��

��

��

Supplementary Figure 1: Plasmids used in this study. Top row is the thresholding sensor: 2 oscillatorplasmids with luxR genes removed and a plasmid containing pArs::luxR. Middle row is the period modulator:2 oscillator plasmids and a plasmid containing pArs::luxI-laa. Bottom row contains 2 plasmids used to studyH2O2 production and synchronization: pLux::ndh and pLux::sodA. NDH-2 synchronization strain is theoscillator plasmids with pZSm45 ndhII.

2

SUPPLEMENTARY INFORMATION

2 | W W W. N A T U R E . C O M / N A T U R E

RESEARCH

Additional Experimental Results

0 200 400 600 800 10000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time (min)

Flu

ore

sce

nce

(a

.u.)

Colony GFP

Mean

Supplementary Figure 2: Biopixels with NDH-2 engineered synchronization observed at ultra-low fluo-rescence (4X, 20ms exposure, 3% power) using an EMCCD camera to ensure no fluorescence interaction.Synchronized oscillations are maintained across the array for the length of the experiment (14 hours).

3


SUPPLEMENTARY INFORMATION RESEARCH

0 500 1000 1500 20000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time (min)

Flu

ore

sce

nce

(a

.u.)

Biopixel GFP

Mean

+ 200 U/mL Catalase

Supplementary Figure 3: Catalase degrades external H2O2 and prevents communication between colonies.When a synchronized population of biopixels was exposed to a step increase of 200 U/ml catalase, syn-chronization was broken and biopixels continued to oscillate individually. Since catalase can’t cross the cellmembrane, this shows that synchronization between colonies depends on H2O2 but oscillations with a colonydo not.

4



RESEARCH

0 50 100 150 2000

0.2

0.4

0.6

0.8

1

Time (min)

Flu

ore

scence (

a.u

.)

Biopixel GFP

Mean

Upstream DownstreamT

ime

Supplementary Figure 4: SodA produces H2O2 internal to the cell, permanently switching the cellularredox state (oxidizing) thereby activating lux-controlled genes. Biopixels rapidly fire and lock on in a spatialwave, far earlier than is typical for colonies of this size. The propagation of ON biopixels suggests thatcolonies are capable of activating those nearby via migrating H2O2 species.

5

Device 1 Device 2

0 50 100 150 200 250 3000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time (min)

Flu

ore

scence (

a.u

.)

Device 1

Device 2

Supplementary Figure 5: Synchronized oscillations occur across 2 fluidically isolated devices held in closeproximity. In this experiment, the devices were started at different times yet become synchronized. Sincethese devices share no common fluid sources or sinks, this confirms that synchronization is mediated byvapor species.

6



Time (min)

Bio

pix

el N

um

be

r

100 200 300 400 500

Supplementary Figure 6: Heatmap of trajectories extracted from low fluorescence intensity control (Suppl.Movie 9) when NDH-2 plasmid is not present. Biopixels oscillate individually but fail to synchronize.

7



RESEARCH

0 100 200 300 400 5000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time (min)

Flu

ore

scence (

a.u

.)+ 25 mM Thiourea

Supplementary Figure 7: The introduction thiourea, a potent radical quencher, produces decaying syn-chronized oscillations across a population of biopixels. Because radical species are precursors for H2O2 ,eliminating them lowers the production of H2O2 and therefore dampens the oscillations. Colonies are stillable to synchronize because, while thiourea eliminates radicals within cells, it does not prevent H2O2 fromdiffusing between cells.

8Supplementary Figure 8: Synchronization is prevented when 100 µg/ml Ampicillin is used in the media.The constructs, strains, and experimental conditions are otherwise identical.

9



Data Analysis

Fluorescence data was obtained by importing fluorescent images into ImageJ and subtracting cellsignal from background signal. Oscillatory period was taken to be the average of peak-to-peakand trough-to-trough distance, calculated using a MATLAB script. The data represented in Fig.1d and 2b-d were collected by stitching 4 images taken at 4X magnification. The mean trajectoryin Fig. 1d was found by averaging 373 individual biopixel trajectories, of which 20 are shown.Biopixel trajectories were extracted from image series using a MATLAB script, where a brightfield image of the corresponding array was used to generate a mask. The data shown in Fig. 2cwas measured over 4 separate experiments using 10-30 oscillatory periods per data point.

Sensor calibration curve (Fig. 2c, bottom) was generated using a series of 2-population ttestscomparing the experimental datasets to randomly generated new sample sets. The mean ofgenerated sets was decremented until the ttest failed with α = 95%, indicating the lowest periodthat could be associated with that arsenite concentration. We repeated this process for eacharsenite level and fit the points with a quadratic since we expected it to take the inverse shape ofthe period vs. arsenite measurements.

Microscopy and Microfluidics

We used a microscopy system similar to our recent studies (1), with the addition of a high-sensitivity Andor DU-897 EMCCD camera. Fluorescent images were taken at 4X every 30 sec-onds using the EMCCD camera (20ms exposure, 97% attentuation) or 2 minutes (2s exposure,90% attenuation) using a standard CCD camera to prevent photobleaching or phototoxicity.

In each device, E. coli cells are loaded from the cell port while keeping the media port at suffi-ciently higher pressure than the waste port below to prevent contamination (Suppl. Fig 8). Cellswere loaded into the cell traps by manually applying pressure pulses to the lines to induce amomentary flow change. The flow was then reversed and allowed for cells to receive fresh mediawith 0.075% Tween which prevented cells from adhering to the main channels and waste ports.

To measure fluid flow rate before each experiment, we measured the streak length of fluorescentbeads (1.0 µm) upon 100 ms exposure to fluorescent light. We averaged at least 1,000 data pointsfor each.

We constructed several microfluidic devices over the course of the study. The trap dimensionswere always 100 µm x 85 µm x 1.65 µm high, which we previously found to be optimal for os-cillator function, except when size was varied to study dynamic interactions. Spacing betweentraps was 25 µm, except in devices designed to study the effects of increasing separation distancebetween traps. For sensor array devices, we constructed 500 and 12,000 trap arrays as well as atandem device which holds two 150 trap arrays in close proximity (25 µm) without sharing fluidsources or sinks.

10



RESEARCH

Media

Waste

Waste

Cell

Supplementary Figure 9: Primary microfluidic device used for this study. Media containing variablearsenite concentration is fed through the cell port, flowing past the biopixel array into the cell and wasteports. During loading, pressure is increased at the cell port and decreased at the waste ports to reverse theflow, allowing cells to pass by the trapping regions. Other microfluidic devices used have the same layoutwith trap number, separation, and size varied.

11



Modeling

To model the dynamics of the quorum-sensing oscillator, we used our previously describedmodel for intracellular concentrations of LuxI (I), AiiA (A), internal AHL (Hi), and external AHL(He) (1),

∂A∂t

= CA[1 − (d/d0)4] G(α, τ) − γA A1 + f (A + I)

(1)

∂I∂t

= CI [1 − (d/d0)4] G(α, τ) − γI I1 + f (A + I)

(2)

∂Hi

∂t=

bI1 + kI

− γH AHi

1 + gA+ D(He − Hi) (3)

∂He

∂t= − d

1 − dD(He − Hi) − µHe + D1

∂2He

∂x2 (4)

In the original model, the concentration of the constitutively produced LuxR protein R wasassumed constant. In the ON/OFF threshold arsenic biosensor circuit, LuxR production is in-duced by arsenic, which we model by the equation

R =αc A

(A0 + A)− γRR (5)

in which the LuxR expression from the arsenic promoter follows a standard saturating functionof the arsenic concentration A. Accordingly, we modified the Hill function for Lux promoter toinclude the explicit dependence on R:

G(α, τ) =δ + α(Rτ Hτ)2

1 + k1(Rτ Hτ)2 (6)

For modeling the period-modulating sensor, we modified the equation for LuxI (2) to includeadditional production from the arsenic promoter,

I = CI [1 − (d/d0)4]G(α, τ) +αc A

(A0 + A)− γI I

(1 + f (A + I)(7)

The following additional parameters were used for the biosensor simulations: αc = 50, A0 =2, γR = .1.

Arsenic levels were swept across the dynamic range of the arsenic promoter to produce thecurve in Fig. 2c. The period for each arsenic level was calculated from the peak-to-peak averageof 15 oscillatory periods.

To model the spatial synchronization of oscillating colonies across a microfluidic array, wegeneralized a simplified “degrade-and-fire" model (6). The delay-differential equation

Xi,j =α(1 + νPi,j,τ2)

(1 +Xi,j,τ1

C0)2

−γXi,j

k + Xi,j(8)

12


1 0 | W W W. N A T U R E . C O M / N A T U R E

RESEARCH

describes oscillations of individual biopixel {i, j} as a combined effect of production and delayedautorepression (first term in the r.h.s.) of the colony-averaged LuxI concentration Xi,j and itsenzymatic degradation by ClpXP (second term). Unlike (6), the first (production) term in Eq. 8describes both delayed auto-repression of LuxI and its delayed activation by H2O2 proportionalto its local concentration Pi,j. Subscripts τ1 and τ2 indicate the delayed concentrations, Xi,j,τ1(t) =Xi,j(t − τ1) and Pi,j,τ2(t) = Pi,j(t − τ2). The dynamics of Pi,j is described by the equation

Pi,j = µ + αpXi,j − γpPi,j + S{Pi,j} (9)

where the first three terms describe the basal and induced production and degradation of H2O2.The last term models the spatial coupling of neighboring biopixels via the H2O2 exchange. For asquare N × N array of traps, we used the following discrete diffusion form of the spatial operator,

S{Pi,j} = D∆−2[Pi−1,j + Pi+1,j + Pi,j−1 + Pi,j+1 − 4Pi,j] (10)

Each colony is affected by the H2O2 produced in four neighboring colonies, two in each dimen-sion of the array, separated by the equal distance ∆. We used the boundary condition Pi,j = 0for the edges of the array i, j = 0, N + 1. This represents the infinite external sink of H2O2 dif-fusing out of the microfluidic chip. The diffusion operator above can be generalized if the rowspacing differs from the column spacing, or for other spatial arrangements of colonies within thebiosensor.

We introduced variability among different traps by randomizing oscillator parameters forindividual traps in each simulation. Specifically, LuxI (X) activation and degradation parameters(p = {α, γ}) of each of the oscillators in the array were varied around their nominal values (p0)as p = p0 + δ where δ is a random number uniformly distributed between −0.25 and 0.25. Weused the following dimensionless parameters for most of our simulations: α0 = 8.25, γ0 = 5.75,ν = 1, τ1 = 10, τ2 = 20, C0 = 6, k = 10, µ = 20, αp = 1, γp = 10, D = 7, ∆ = 1.

For the characterization of various regimes of array synchronization, 16 colonies were mod-eled in the 4 × 4 array. Scaling up the simulation with larger numbers of colonies producedequivalent results. Overproduction of H2O2 by expressing sodA was captured by increasing αp

20-fold. This is consistent with expression from a pSC101m plasmid with a copy number of20-30. Depletion of external H2O2 by catalase was modeled by increasing H2O2 degradation (γp)and decreasing H2O2 diffusion, D. In Suppl. Fig. 9 we show the variance of the concentrationsXi,j within the array averaged over time and parameter variations. This plot demonstrates thatthe synchronicity among the biopixels decreases with increase of spacing among them, and for∆ > 5 is completely lost.

Increasing the trap spacing ∆ 2-fold while simultaneously decreasing k 4-fold allowed usto reproduce the more complex waveforms observed experimentally in our arrays. Note thatchanging k models the change of the trap depth. As the size of the trap decreases, the flowof media is able to more rapidly sweep away AHL and increase the effective degradation forthe colony. Simulating smaller and more sparse trap sizes recovered antiphase behavior forneighboring biopixels (Suppl. Fig. 10). We also simulated the arrays with traps of two differentsizes in different rows and recovered the experimental 2:1 biopixel resonance or 2:1 + antiphasebehavior depending on the trap spacing (Fig. 3d, bottom).

13

W W W. N A T U R E . C O M / N A T U R E | 1 1


0 5 10 15 204

6

8

10

12

14

16

18

20Average Simulation Variance

Trap Distance (A.U.)

Variance (

A.U

.)

Supplementary Figure 10: Computational results depicting biopixel synchronicity as a function of trapseparation distance. As biopixels are moved farther apart, the entropy increases due to decreased effectivemigration of H2O2 between colonies.

The model was also able to capture the alternating large and small amplitude oscillationsobserved in the ON/OFF biosensor (Suppl. Fig. 11). This behavior was seen when C0 was in-creased 2-fold, capturing the decreased level of LuxR in ON/OFF experiments where it was thelimiting factor for oscillations.

14


1 2 | W W W. N A T U R E . C O M / N A T U R E

RESEARCH

0 100 200 300 400 500 600 700 800 9000

50

100

150

200

250

300

350

Flu

ore

sce

nc

e (

a.u

.)

Time (min)

Supplementary Figure 11: Antiphase behavior of 4 neighboring biopixels having equal trap sizes andspacing ∆ = 3.

15

W W W. N A T U R E . C O M / N A T U R E | 1 3


0 100 200 300 400 5000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Time (min)

Flu

ore

scence (

a.u

.)

0 100 200 5000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Flu

ore

scence (

a.u

.)

Time (min)300 400

Experiment -- limited LuxR (pArs::LuxR @ 1.6 uM arsenite) Model-- constant low LuxR

Supplementary Figure 12: Oscillations of alternating large and small amplitude when LuxR is limited inexperiments and simulations. The alternating oscillations vanish when LuxR is restored to its normal levelin the model. Experimentally, we were unable to build a system in which LuxR is tunable between big/smalland normal amplitude regimes. This is probably due to the small dynamic range of arsenite promoter-drivenoutput of LuxR compared to the level produced by 3 constitutively expressed copies in the original circuit.

16

References

[1] Danino, T., Mondragon-Palomino, O., Tsimring, L. & Hasty, J. A synchronized quorum ofgenetic clocks. Nature 463, 326–330 (2010).

[2] Pedelacq, J.-D., Cabantous, S., Tran, T., Terwilliger, T. C. & Waldo, G. S. Engineering andcharacterization of a superfolder green fluorescent protein. Nature Biotechnology 24, 79–88(2006).

[3] Stocker, J. et al. Development of a set of simple bacterial biosensors for quantitative and rapidmeasurements of arsenite and arsenate in potable water. Environ. Sci. Technol 37, 4743–4750(2003).

[4] Keiler, K., Waller, P. & Sauer, R. Role of a peptide tagging system in degradation of proteinssynthesized from damaged messenger rna. Science 271, 990 (1996).

[5] Quan, J. & Tian, J. Circular polymerase extension cloning of complex gene libraries andpathways. PloS one 4, e6441 (2009).

[6] Mather, W., Bennett, M., Hasty, J. & Tsimring, L. Delay-induced degrade-and-fire oscillationsin small genetic circuits. Physical Review Letters 102, 068105 (2009).

17

a sensing array of radically coupled genetic 'biopixels' · 2017-08-22 · a sensing...

Documents