Synthetic Biology - WordPress.com · Synthetic biology is the engineering of biology: the deliberate (re)design and construction of novel biological and biologically based parts,

Gabriele Lillacci, Ph.D. !

ETH Zürich Department of Biosystems Science and Engineering

Synthetic Biology

What is Synthetic Biology?

!2


!2


!2


Synthetic biology is the engineering of biology:

the deliberate (re)design and construction of

novel biological and biologically based parts,

devices and systems to perform new functions

for useful purposes, that draws on principles

elucidated from biology and engineering.

!3

Programming Biology

!4

Programming Biology

!4

Programming Biology

Gene

!4

Programming Biology

Gene

mRNA

!4

Programming Biology

Gene

mRNA

Protein

!4

Programming Biology

Gene

mRNA

Protein

Gene Expression

!4

Programming Biology

Gene

mRNA

Protein

Gene Expression Genetic Engineering

!4

Programming Biology

Synthetic Biology

!5

Programming Biology

Synthetic Biology

!5

Programming Biology

Synthetic Biology

!5

Programming Biology

Synthetic Biology

!5

Programming Biology

Synthetic Biology

!5

• Artemisinin

!

• Biocomputer

!

• Cyborg Yeast

Three Stories

!6

Artemisinin

!7

• Artemisinin is a drug against malaria.

Artemisinin

!7


• It comes from a plant, Artemisia annua.

Artemisinin

!7


• It comes from a plant, Artemisia annua.

• The supply of plant-based artemisinin is unstable, leading to shortages and price fluctuations.

Artemisinin

!7

LETTERdoi:10.1038/nature12051

High-level semi-synthetic production of the potentantimalarial artemisininC. J. Paddon1, P. J. Westfall1{, D. J. Pitera1, K. Benjamin1, K. Fisher1, D. McPhee1, M. D. Leavell1, A. Tai1, A. Main1{, D. Eng1,D. R. Polichuk2, K. H. Teoh2{, D. W. Reed2, T. Treynor1, J. Lenihan1{, M. Fleck1, S. Bajad1{, G. Dang1{, D. Dengrove1, D. Diola1,G. Dorin1, K. W. Ellens2{, S. Fickes1, J. Galazzo1, S. P. Gaucher1, T. Geistlinger1, R. Henry1, M. Hepp2{, T. Horning1, T. Iqbal1,H. Jiang1, L. Kizer1, B. Lieu1, D. Melis1, N. Moss1, R. Regentin1{, S. Secrest1, H. Tsuruta1, R. Vazquez1, L. F. Westblade1, L. Xu1, M. Yu1,Y. Zhang2{, L. Zhao1, J. Lievense1{, P. S. Covello2, J. D. Keasling3,4,5,6, K. K. Reiling1, N. S. Renninger1 & J. D. Newman1

In 2010 there were more than 200 million cases of malaria, and atleast 655,000 deaths1. The World Health Organization has recom-mended artemisinin-based combination therapies (ACTs) for thetreatment of uncomplicated malaria caused by the parasitePlasmodium falciparum. Artemisinin is a sesquiterpene endoper-oxide with potent antimalarial properties, produced by the plantArtemisia annua. However, the supply of plant-derived artemisi-nin is unstable, resulting in shortages and price fluctuations, com-plicating production planning by ACT manufacturers2. A stablesource of affordable artemisinin is required. Here we use syntheticbiology to develop strains of Saccharomyces cerevisiae (baker’syeast) for high-yielding biological production of artemisinic acid,a precursor of artemisinin. Previous attempts to produce commer-cially relevant concentrations of artemisinic acid were unsuccess-ful, allowing production of only 1.6 grams per litre of artemisinicacid3. Here we demonstrate the complete biosynthetic pathway,including the discovery of a plant dehydrogenase and a secondcytochrome that provide an efficient biosynthetic route to artemi-sinic acid, with fermentation titres of 25 grams per litre of artemi-sinic acid. Furthermore, we have developed a practical, efficientand scalable chemical process for the conversion of artemisinicacid to artemisinin using a chemical source of singlet oxygen, thusavoiding the need for specialized photochemical equipment. Thestrains and processes described here form the basis of a viableindustrial process for the production of semi-synthetic artemisininto stabilize the supply of artemisinin for derivatization into activepharmaceutical ingredients (for example, artesunate) for incor-poration into ACTs. Because all intellectual property rights havebeen provided free of charge, this technology has the potential toincrease provision of first-line antimalarial treatments to the deve-loping world at a reduced average annual price.

Before the discovery of the enzymes that complete the biosyntheticpathway of artemisinin production (see Supplementary Fig. 1 for acomplete overview), several improvements were made to the originalamorphadiene-producing strain Y337 (ref. 3). We replaced the MET3promoter with the copper-regulated CTR3 promoter (Fig. 1a), enab-ling restriction of ERG9 expression (ERG9 encodes squalene syn-thase, which catalyses the competing reaction of joining two farnesyldiphosphate moieties to form squalene) by addition of the inexpen-sive repressor CuSO4 to the medium rather than the more expensivemethionine4–6. Strains Y1516 (PCTR3-ERG9) and Y337 (PMET3-ERG9)

(Supplementary Table 1) both produced similar amounts of amorpha-diene (Supplementary Fig. 2), demonstrating the equivalence of theMET3 and CTR3 promoters for repression of ERG9 expression. Wecompared the production of amorphadiene from Y337 with the pro-duction of artemisinic acid from Y285, a variant of Y337 that alsoexpressed the amorphadiene oxidase CYP71AV1 (a cytochromeP450) and A. annua CPR1 (its cognate reductase) from a high-copyplasmid (pAM322)3. Both strains were grown in a fed-batch fermentorwith mixed glucose and ethanol feed. Whereas Y337 produced morethan 12 g l21 of amorphadiene, Y285 produced significantly less sesqui-terpene: 3.3 g l21 of artemisinic acid (Fig. 2a and Supplementary Table2) plus 0.3 g l21 amorphadiene, 0.18 g l21 artemisinic alcohol and nodetectable artemisinic aldehyde (Supplementary Table 3). The viabilityof the Y285 culture also decreased markedly after CYP71AV1 and CPR1expression (Fig. 2a). We surmised that the decreased viability andreduced production of sesquiterpene products in Y285 might be causedby the cytochrome P450 responsible for oxidizing amorphadiene, or bythe rapid accumulation of artemisinic acid.

Poor coupling between P450 cytochromes and their reductases canresult in the release of reactive oxygen species7. In liver microsomes,the P450 enzyme is generally present in excess over its reductase8,whereas in Y285 and Y301 both enzymes are expressed from stronggalactose-regulated promoters on a high-copy plasmid, and are pre-sumably present at similar levels. We reduced expression of CPR1 byexpressing it from a weaker promoter (GAL3 promoter) and integrat-ing a single copy into genomic DNA, generating strain Y657 (Sup-plementary Table 1). Y657 had an increase in cell growth (Fig. 2b) andviability (Supplementary Fig. 3) compared to either Y285 or Y301(isogenic to Y285, but with PCTR3-ERG9 replacing PMET3-ERG9), butshowed lower artemisinic acid production in shake-flask cultures(Fig. 2c) and mixed-feed fed-batch fermentors (SupplementaryTables 2 and 3). Comparison of all amorphadiene-derived sesquiter-penes showed that although reducing CPR1 expression decreased arte-misinic acid production, total sesquiterpene production remainedrelatively high, indicating that low CPR1 levels increase cell health,but decrease the total rate of amorphadiene oxidations (Fig. 3a; com-pare Y301 and Y657). The reaction rate of some cytochromes P450 isenhanced by their interaction with cytochrome b5 as explained byseveral possible mechanisms9,10. We identified a cytochrome b5 com-plementary DNA from A. annua (CYB5; Supplementary Fig. 4) andexpressed a chromosomally integrated copy from a strong promoter

1Amyris, Inc., 5885 Hollis Street, Suite 100, Emeryville, California 94608, USA. 2National Research Council of Canada, 110 Gymnasium Place, Saskatoon, Saskatchewan S7N 0W9, Canada. 3Department ofChemical and Biomolecular Engineering, University of California, Berkeley, California 94720, USA. 4Department of Bioengineering, University of California, Berkeley, California 94720, USA. 5PhysicalBiosciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA. 6Joint BioEnergy Institute, 5885 Hollis Street, Emeryville, California 94608, USA. {Present addresses: DowAgroSciences, 9330 Zionsville Road, Indianapolis, Indiana 46268, USA (P.J.W.); Radiant Genomics, Inc., 2430 Fifth Street, Suite D, Berkeley, California 94710, USA (A.M.); Arkansas Biosciences Institute,Arkansas State University, Jonesboro, Arkansas 72401, USA (K.H.T.); Solazyme, Inc., 225 Gateway Boulevard, South San Francisco, California 94080, USA (J. Lenihan); Sutro Biopharma Inc., 310 UtahAvenue, South San Francisco, California 94080, USA (S.B.); Abbvie, 1500 Seaport Boulevard, Redwood City, California 94063, USA (G. Dang); Horticultural Sciences Department, University of Florida, POBox 110690, Gainesville, Florida 32611-0690, USA (K.W.E.); Department of Plant Sciences, University of Saskatchewan, 51 Campus Drive, Saskatoon, Saskatchewan S7N 5A8, Canada (M.H.); RegentinConsulting, 25409 Modoc Court, Hayward, California 94542, USA (R.R.); AS Key Laboratory of Plant Germplasm Enhancement and Specialty Agriculture, Wuhan Botanical Garden, The Chinese Academy ofScience, Wuhan 430074, China (Y.Z.); Genomatica, Inc., 10520 Wateridge Circle, San Diego, California 92121, USA (J. Lievense).

5 2 8 | N A T U R E | V O L 4 9 6 | 2 5 A P R I L 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

The Breakthrough• Thanks to a synthetic gene network implemented in

baking yeast, large-scale industrial production of artemisinin is now possible.

!8

LETTERdoi:10.1038/nature12051

High-level semi-synthetic production of the potentantimalarial artemisininC. J. Paddon1, P. J. Westfall1{, D. J. Pitera1, K. Benjamin1, K. Fisher1, D. McPhee1, M. D. Leavell1, A. Tai1, A. Main1{, D. Eng1,D. R. Polichuk2, K. H. Teoh2{, D. W. Reed2, T. Treynor1, J. Lenihan1{, M. Fleck1, S. Bajad1{, G. Dang1{, D. Dengrove1, D. Diola1,G. Dorin1, K. W. Ellens2{, S. Fickes1, J. Galazzo1, S. P. Gaucher1, T. Geistlinger1, R. Henry1, M. Hepp2{, T. Horning1, T. Iqbal1,H. Jiang1, L. Kizer1, B. Lieu1, D. Melis1, N. Moss1, R. Regentin1{, S. Secrest1, H. Tsuruta1, R. Vazquez1, L. F. Westblade1, L. Xu1, M. Yu1,Y. Zhang2{, L. Zhao1, J. Lievense1{, P. S. Covello2, J. D. Keasling3,4,5,6, K. K. Reiling1, N. S. Renninger1 & J. D. Newman1

In 2010 there were more than 200 million cases of malaria, and atleast 655,000 deaths1. The World Health Organization has recom-mended artemisinin-based combination therapies (ACTs) for thetreatment of uncomplicated malaria caused by the parasitePlasmodium falciparum. Artemisinin is a sesquiterpene endoper-oxide with potent antimalarial properties, produced by the plantArtemisia annua. However, the supply of plant-derived artemisi-nin is unstable, resulting in shortages and price fluctuations, com-plicating production planning by ACT manufacturers2. A stablesource of affordable artemisinin is required. Here we use syntheticbiology to develop strains of Saccharomyces cerevisiae (baker’syeast) for high-yielding biological production of artemisinic acid,a precursor of artemisinin. Previous attempts to produce commer-cially relevant concentrations of artemisinic acid were unsuccess-ful, allowing production of only 1.6 grams per litre of artemisinicacid3. Here we demonstrate the complete biosynthetic pathway,including the discovery of a plant dehydrogenase and a secondcytochrome that provide an efficient biosynthetic route to artemi-sinic acid, with fermentation titres of 25 grams per litre of artemi-sinic acid. Furthermore, we have developed a practical, efficientand scalable chemical process for the conversion of artemisinicacid to artemisinin using a chemical source of singlet oxygen, thusavoiding the need for specialized photochemical equipment. Thestrains and processes described here form the basis of a viableindustrial process for the production of semi-synthetic artemisininto stabilize the supply of artemisinin for derivatization into activepharmaceutical ingredients (for example, artesunate) for incor-poration into ACTs. Because all intellectual property rights havebeen provided free of charge, this technology has the potential toincrease provision of first-line antimalarial treatments to the deve-loping world at a reduced average annual price.

Before the discovery of the enzymes that complete the biosyntheticpathway of artemisinin production (see Supplementary Fig. 1 for acomplete overview), several improvements were made to the originalamorphadiene-producing strain Y337 (ref. 3). We replaced the MET3promoter with the copper-regulated CTR3 promoter (Fig. 1a), enab-ling restriction of ERG9 expression (ERG9 encodes squalene syn-thase, which catalyses the competing reaction of joining two farnesyldiphosphate moieties to form squalene) by addition of the inexpen-sive repressor CuSO4 to the medium rather than the more expensivemethionine4–6. Strains Y1516 (PCTR3-ERG9) and Y337 (PMET3-ERG9)

(Supplementary Table 1) both produced similar amounts of amorpha-diene (Supplementary Fig. 2), demonstrating the equivalence of theMET3 and CTR3 promoters for repression of ERG9 expression. Wecompared the production of amorphadiene from Y337 with the pro-duction of artemisinic acid from Y285, a variant of Y337 that alsoexpressed the amorphadiene oxidase CYP71AV1 (a cytochromeP450) and A. annua CPR1 (its cognate reductase) from a high-copyplasmid (pAM322)3. Both strains were grown in a fed-batch fermentorwith mixed glucose and ethanol feed. Whereas Y337 produced morethan 12 g l21 of amorphadiene, Y285 produced significantly less sesqui-terpene: 3.3 g l21 of artemisinic acid (Fig. 2a and Supplementary Table2) plus 0.3 g l21 amorphadiene, 0.18 g l21 artemisinic alcohol and nodetectable artemisinic aldehyde (Supplementary Table 3). The viabilityof the Y285 culture also decreased markedly after CYP71AV1 and CPR1expression (Fig. 2a). We surmised that the decreased viability andreduced production of sesquiterpene products in Y285 might be causedby the cytochrome P450 responsible for oxidizing amorphadiene, or bythe rapid accumulation of artemisinic acid.

Poor coupling between P450 cytochromes and their reductases canresult in the release of reactive oxygen species7. In liver microsomes,the P450 enzyme is generally present in excess over its reductase8,whereas in Y285 and Y301 both enzymes are expressed from stronggalactose-regulated promoters on a high-copy plasmid, and are pre-sumably present at similar levels. We reduced expression of CPR1 byexpressing it from a weaker promoter (GAL3 promoter) and integrat-ing a single copy into genomic DNA, generating strain Y657 (Sup-plementary Table 1). Y657 had an increase in cell growth (Fig. 2b) andviability (Supplementary Fig. 3) compared to either Y285 or Y301(isogenic to Y285, but with PCTR3-ERG9 replacing PMET3-ERG9), butshowed lower artemisinic acid production in shake-flask cultures(Fig. 2c) and mixed-feed fed-batch fermentors (SupplementaryTables 2 and 3). Comparison of all amorphadiene-derived sesquiter-penes showed that although reducing CPR1 expression decreased arte-misinic acid production, total sesquiterpene production remainedrelatively high, indicating that low CPR1 levels increase cell health,but decrease the total rate of amorphadiene oxidations (Fig. 3a; com-pare Y301 and Y657). The reaction rate of some cytochromes P450 isenhanced by their interaction with cytochrome b5 as explained byseveral possible mechanisms9,10. We identified a cytochrome b5 com-plementary DNA from A. annua (CYB5; Supplementary Fig. 4) andexpressed a chromosomally integrated copy from a strong promoter

1Amyris, Inc., 5885 Hollis Street, Suite 100, Emeryville, California 94608, USA. 2National Research Council of Canada, 110 Gymnasium Place, Saskatoon, Saskatchewan S7N 0W9, Canada. 3Department ofChemical and Biomolecular Engineering, University of California, Berkeley, California 94720, USA. 4Department of Bioengineering, University of California, Berkeley, California 94720, USA. 5PhysicalBiosciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA. 6Joint BioEnergy Institute, 5885 Hollis Street, Emeryville, California 94608, USA. {Present addresses: DowAgroSciences, 9330 Zionsville Road, Indianapolis, Indiana 46268, USA (P.J.W.); Radiant Genomics, Inc., 2430 Fifth Street, Suite D, Berkeley, California 94710, USA (A.M.); Arkansas Biosciences Institute,Arkansas State University, Jonesboro, Arkansas 72401, USA (K.H.T.); Solazyme, Inc., 225 Gateway Boulevard, South San Francisco, California 94080, USA (J. Lenihan); Sutro Biopharma Inc., 310 UtahAvenue, South San Francisco, California 94080, USA (S.B.); Abbvie, 1500 Seaport Boulevard, Redwood City, California 94063, USA (G. Dang); Horticultural Sciences Department, University of Florida, POBox 110690, Gainesville, Florida 32611-0690, USA (K.W.E.); Department of Plant Sciences, University of Saskatchewan, 51 Campus Drive, Saskatoon, Saskatchewan S7N 5A8, Canada (M.H.); RegentinConsulting, 25409 Modoc Court, Hayward, California 94542, USA (R.R.); AS Key Laboratory of Plant Germplasm Enhancement and Specialty Agriculture, Wuhan Botanical Garden, The Chinese Academy ofScience, Wuhan 430074, China (Y.Z.); Genomatica, Inc., 10520 Wateridge Circle, San Diego, California 92121, USA (J. Lievense).

5 2 8 | N A T U R E | V O L 4 9 6 | 2 5 A P R I L 2 0 1 3


What It Takes

(GAL7 promoter) in a strain with low CPR1 expression. The resultingstrain, Y692 (Supplementary Table 1), produced higher concentrationsof artemisinic acid than strains without CYB5 (Fig. 2c and Sup-plementary Tables 2 and 3; compare Y657 and Y692). Expression ofCYB5 also increased the production of artemisinic aldehyde, leading toa 40% increase in total sesquiterpene production in shake-flaskcultures (Fig. 3a; compare Y657 and Y692), and almost doubled theproduction of artemisinic aldehyde in fermentors (SupplementaryTable 3). In view of the reactivity and presumed toxicity of artemisinicaldehyde, we expressed a recently isolated cDNA encoding A. annuaartemisinic aldehyde dehydrogenase (ALDH1)11 in Y692 to produceY973 and Y1368 (also expressing increased levels of cytosolic catalaseto reduce oxidative stress; Supplementary Table 1). Expression ofALDH1 markedly increased the production of artemisinic acid in bothflask (Fig. 2c; Y1368) and fermentor cultures (Supplementary Tables 2and 3; Y973). Artemisinic aldehyde was undetectable in flask cultures(Fig. 3a; Y1368), and barely detectable in fermentors (Supplemen-tary Table 3; Y973). Furthermore, the expression of ALDH1 in Y973allowed early induction of fermentor cultures immediately after inocu-lation (previous attempts at early induction with Y285 and Y301 hadresulted in rapid loss of viability), further increasing production to7.7 g l21 artemisinic acid (Supplementary Tables 2 and 4). The yield(Cmol% of substrate carbon incorporated into artemisinic acid) wasmore than doubled compared to the initial Y285 cultures (Supplemen-tary Tables 3 and 4).

In the course of investigating the biosynthesis of artemisinin inA. annua glandular trichomes, a gene encoding a putative alcoholdehydrogenase (ADH1) was examined. The gene is represented by acontiguous set of glandular trichome-derived expressed sequence tags(ESTs)12 corresponding to 1.3% of the EST collection. The A. annuaADH1 open reading frame (ORF) was expressed as a fusion proteinand purified from Escherichia coli. Sequence analysis and in vitrocharacterization revealed that ADH1 is an NAD-dependent alcoholdehydrogenase of the medium chain dehydrogenase/reductase super-family, with specificity towards artemisinic alcohol (Michaelis constant(Km) 5 11 6 3mM, kcat 5 41 6 5 s21 (mean 6 s.e.m.); SupplementaryFig. 5). This specificity and the evidence for strong glandular trichomeexpression indicate a role for ADH1 in the formation of artemisinicaldehyde in the artemisinin pathway of A. annua. Therefore, we pro-pose that all five enzymes (CYP71AV1, CPR1, CYB5, ADH1 andALDH1) are involved in the oxidation of amorphadiene to artemisinicacid in A. annua plants, and set out to reconstitute the entire hetero-logous biosynthetic pathway in yeast (Fig. 1b).

Observing the accumulation of artemisinic alcohol in strain Y1368(Fig. 3a), we completed the biosynthetic pathway by expressing ADH1in conjunction with ALDH1, CYP71AV1, CYB5 and CPR1. The result-ing strain, Y1283, produced no detectable artemisinic alcohol in flaskcultures, and increased artemisinic acid production by 18% (Figs 2cand 3a; compare Y1368 and Y1283). In fermentors, ADH1 increasedproduction to 8.1 g l21 in an early induction, mixed-feed process, while

OPP

CH3 SCoA

O

Acetyl-CoA

ERG10ERG13

tHMG1 (X3)OH

HO2CCH3 OH

Mevalonate

ERG12ERG8

ERG19

CH3

OPPCH2

CH3

CH3 OPP

ERG20

P -ERG9CTR3

Ergosterol

ADS

OHH H

ADH1ALDH1

ERG1 ERG7 ERG11 ERG24 ERG2 ERG25 ERG6 ERG2 ERG3 ERG5 ERG4

IPP DMAPP

FPP

H

H

O

H

Artemisinic acid

O

P -ERG9MET3

IDI1

H

HOH

H

HH

O

H

H

O

H O

H

H

ADH1

ALDH1

Artemisinic acid

Artemisinic aldehyde

Artemisinic alcohol

CYP71AV1, CPR1, CYB5

CYP71AV1, CPR1, CYB5

a b Figure 1 | Artemisinic acid production pathwayin S. cerevisiae and summary of strains described.a, Overview of artemisinic acid productionpathway. Overexpressed genes controlled by theGAL induction system are shown in green.Copper- or methionine-repressed squalenesynthase (ERG9) is shown in red. DMAPP,dimethylallyl diphosphate; FPP, farnesyldiphosphate; IPP, isopentenyl diphosphate.tHMG1 encodes truncated HMG-CoA reductase.b, The full three-step oxidation of amorphadiene toartemisinic acid from A. annua expressed in S.cerevisiae. CYP71AV1, CPR1 and CYB5 oxidizeamorphadiene to artemisinic alcohol; ADH1oxidizes artemisinic alcohol to artemisinicaldehyde; ALDH1 oxidizes artemisinic aldehyde toartemisinic acid. Strains containing these genes aredescribed in Supplementary Table 1.

0

10

20

30

40

50

60

70

80

90

100

0

2

4

6

8

10

12

14

0 20 40 60 80 100

Art

emis

inic

aci

d, a

mor

phad

iene

(g l–1

)

Time (h)

Y285 AAY337 ADY285 viabilityY337 viability

Cell viability (%

viable)

a

0

4

8

12

16

0 24 48 72

Gro

wth

(D60

0 nm

)

Time (h)

Y285

Y301

Y657

b

0

0.2

0.4

0.6

0.8

24 48 72

Art

emis

inic

aci

d (g

l–1)

Time (h)

Y285Y301Y657Y692Y1368Y1283

c

Figure 2 | Growth, viability and production by S. cerevisiae strains.a, Production and cell viability of artemisinic acid production strain Y285 andamorphadiene production strain Y337 in the glucose and ethanol mixed-feed,fed-batch fermentation process. AA, artemisinic acid; AD, amorphadiene.

b, Growth of artemisinic-acid-producing strains in shake-flasks. c, Productionof artemisinic acid in shake-flasks by different strains. Error bars denotestandard deviation of triplicate shake-flask cultures.

LETTER RESEARCH

2 5 A P R I L 2 0 1 3 | V O L 4 9 6 | N A T U R E | 5 2 9


!9

Biocomputer

!10

Biocomputer• micro-RNA are small fragments of RNA that act as

regulators of gene expression.

!10



• Different types of cells have different micro-RNA: they are a signature of cell identity.

!10



• Different types of cells have different micro-RNA: they are a signature of cell identity.

• Cancer cells also have different micro-RNA signatures compared to the tissues from which they originated.

!10

Biocomputer• Can cancer cells be detected based on their

micro-RNA signature?

!11

knockdown ES cells had little 5caC excisionactivity (Fig. 4D). Moreover, immunodepletionof TDG from the ES cell nuclear extract greatlyreduced the 5caC excision activity (Fig. 4A,lane 3). These results indicate that TDG is ableto recognize and excise 5caC, an oxidationproduct of 5mC, in duplex DNA.

Stable ES cell lines expressing a Tdg-specificsmall interfering RNA were established, andTDG depletion was confirmed by Western an-alysis (fig. S12). By using triple quadrupole massspectrometry, we could detect 5caC in genomicDNA isolated from TDG-depleted ES cells, butno reliable signal was detected in TDG-proficientcontrol cells expressing scramble short hairpinRNA (shRNA) (Fig. 4E). Similarly, 5caC wasdetectable in mouse induced pluripotent stem(iPS) cells when the Tdg gene was knocked out(fig. S13). Judging from our calculation basedon the measurement of a 5caC standard, the num-ber of 5caC per genome is ~9000 in Tdg-depletedES or iPS cells but below 1000 in wild-type cells.

TDG has been implicated in DNA demeth-ylation for its function in excising the deami-nation product of 5mC, 5hmC, or 5mC itselffrom DNA (17–19), yet mammalian TDG lacksglycosylase activity toward 5mC (6, 12). Al-though TDG is able to excise 5hmU (19), thedeamination product of 5hmC, our work pro-vides evidence that the Tet dioxygenases oxi-dize 5mC and 5hmC to 5caC, which becomesa substrate for TDG. Therefore, Tet-mediatedconversion of 5mC and 5hmC to 5caC couldtrigger TDG-initiated BER, as indicated here.These sequential events would lead to DNAdemethylation, because unmethylated cytosinesare inserted into the repaired genomic region(fig. S14).

Genome-wide mapping revealed that Tet1 isrelatively enriched in CpG-rich active promotersthat are unmethylated (20–23), but 5hmC is un-derrepresented in the majority of Tet1 bindingsites in ES cells (24–26). These apparent para-doxes might be accounted for if active pro-moters with Tet1 binding sites were preventedfrom erroneous hypermethylation because of Tet1oxidizing 5mC into 5caC, which could then beremoved by TDG-mediated BER repair. In thiscase, 5mC is most likely undetectable in the ac-tive promoters because of their transient exis-tence in a small proportion of cells. Likewise, inmany of the Tet1 binding sites, 5hmC could beunderrepresented because of conversion to 5caC,which is rapidly removed in cells.

Note added in proof: During the revisionof this manuscript, Ito et al.’s report (www.sciencemag.org/content/early/2011/07/20/science.1210597.abstract) appeared online de-scribing the enzymatic activity of Tet proteinsin the conversion of 5mC to 5fC and 5caC, aswell as the detection of these derivatives in mousegenomic DNA.

References and Notes1. R. Jaenisch, A. Bird, Nat. Genet. 33 (suppl.), 245

(2003).2. S. Simonsson, J. Gurdon, Nat. Cell Biol. 6, 984

(2004).3. X. J. He, T. Chen, J. K. Zhu, Cell Res. 21, 442

(2011).4. Z. Liutkeviciute, G. Lukinavicius, V. Masevicius,

D. Daujotyte, S. Klimasauskas, Nat. Chem. Biol. 5, 400(2009).

5. C. P. Walsh, G. L. Xu, Curr. Top. Microbiol. Immunol.301, 283 (2006).

6. S. C. Wu, Y. Zhang, Nat. Rev. Mol. Cell Biol. 11, 607(2010).

7. C. Dahl, K. Grønbæk, P. Guldberg, Clin. Chim. Acta 412,831 (2011).

8. M. Tahiliani et al., Science 324, 930 (2009).9. T. Pfaffeneder et al., Angew. Chem. Int. Ed. Engl. 50,

7008 (2011).10. D. Globisch et al., PLoS ONE 5, e15367 (2010).11. T. Lindahl, R. D. Wood, Science 286, 1897 (1999).12. D. Cortázar et al., Nature 470, 419 (2011).13. M. T. Bennett et al., J. Am. Chem. Soc. 128, 12510

(2006).14. B. Hendrich, U. Hardeland, H. H. Ng, J. Jiricny, A. Bird,

Nature 401, 301 (1999).15. H. E. Krokan, R. Standal, G. Slupphaug, Biochem. J. 325,

1 (1997).16. R. J. Boorstein et al., J. Biol. Chem. 276, 41991

(2001).17. R. Métivier et al., Nature 452, 45 (2008).18. B. Zhu et al., Proc. Natl. Acad. Sci. U.S.A. 97, 5135

(2000).19. S. Cortellino et al., Cell 146, 67 (2011).20. W. A. Pastor et al., Nature 473, 394 (2011).21. G. Ficz et al., Nature 473, 398 (2011).22. C. X. Song et al., Nat. Biotechnol. 29, 68 (2011).23. H. Wu et al., Genes Dev. 25, 679 (2011).24. K. Williams et al., Nature 473, 343 (2011).25. Y. Xu et al., Mol. Cell 42, 451 (2011).26. H. Wu et al., Nature 473, 389 (2011).Acknowledgments: We thank C. Walsh for critical reading

of the manuscript, G. Shi and S. Klimasauskas fordiscussions, J. Ju for providing Tet cDNA clones, T. Carellfor 2′-deoxy-5-carboxylcytidine and Z. Hua for the TDGantibody. This study was supported by grants from theMinistry of Science and Technology China (2007CB947503and 2009CB941101 to G.-L.X., 2010CB912100 to L.L.),National Science Foundation of China (30730059 toG.-L.X., 30930052 and 30821065 to L.L.), and theStrategic Priority Research Program of the Chinese Academyof Sciences (XDA01010301 to G.-L.X.) and by the NIH(GM071440 to C.H.) and (1S10RR027643-01 to K.Z.).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1210944/DC1Materials and MethodsFigs. S1 to S14References (27–32)

11 July 2011; accepted 25 July 2011Published online 4 August 2011;10.1126/science.1210944

Multi-Input RNAi-Based Logic Circuitfor Identification of SpecificCancer CellsZhen Xie,1,2* Liliana Wroblewska,2 Laura Prochazka,3 Ron Weiss,2,4† Yaakov Benenson1,3†‡

Engineered biological systems that integrate multi-input sensing, sophisticated informationprocessing, and precisely regulated actuation in living cells could be useful in a variety ofapplications. For example, anticancer therapies could be engineered to detect and respond tocomplex cellular conditions in individual cells with high specificity. Here, we show a scalabletranscriptional/posttranscriptional synthetic regulatory circuit—a cell-type “classifier”—that sensesexpression levels of a customizable set of endogenous microRNAs and triggers a cellular responseonly if the expression levels match a predetermined profile of interest. We demonstrate that a HeLacancer cell classifier selectively identifies HeLa cells and triggers apoptosis without affectingnon-HeLa cell types. This approach also provides a general platform for programmed responsesto other complex cell states.

Synthetic biomolecular pathways with elab-orate information processing capabilitywill enable in situ response to complex

physiological conditions (1). For example, mul-

tiple cancer-specific biomarkers (2) can be sensedand integrated in such pathways, providing pre-cise control of therapeutic agents (3). A combina-tion of up to two tissue-specific signals, including

promoter and/or microRNA (miRNA) activity,mRNA, and protein levels, have been used topartially restrict therapeutic action to cancer cells(4–9). In parallel, research in synthetic biology hasdemonstrated multi-input information processingin living cells (10–18), but the interaction be-tween these systems and the cellular context hasbeen limited. Yet, general-purpose mechanismsfor programmable integration of multiple mark-ers are required to detect cell state and precisely

1Faculty of Arts and Sciences (FAS) Center for SystemsBiology, Harvard University, 52 Oxford Street, Cambridge,MA 02138, USA. 2Department of Biological Engineering,Massachusetts Institute of Technology (MIT), 40 Ames Street,Cambridge, MA 02142, USA. 3Department of Biosystems Sci-ence and Engineering, Eidgenössische Technische Hochschule(ETH) Zürich, Mattenstrasse 26, Basel 4058, Switzerland. 4De-partment of Electrical Engineering and Computer Science, MIT,40 Ames Street, Cambridge MA 02142, USA.

*Present address: Department of Biological Engineering,MIT, 40 Ames Street, Cambridge, MA 02142, USA.†To whom correspondence should be addressed. E-mail:[email protected] (Y.B.); [email protected] (R.W.)‡Present address: Department of Biosystems Science andEngineering, ETH Zürich, Mattenstrasse 26, Basel 4058,Switzerland.

www.sciencemag.org SCIENCE VOL 333 2 SEPTEMBER 2011 1307

REPORTS

knockdown ES cells had little 5caC excisionactivity (Fig. 4D). Moreover, immunodepletionof TDG from the ES cell nuclear extract greatlyreduced the 5caC excision activity (Fig. 4A,lane 3). These results indicate that TDG is ableto recognize and excise 5caC, an oxidationproduct of 5mC, in duplex DNA.

Stable ES cell lines expressing a Tdg-specificsmall interfering RNA were established, andTDG depletion was confirmed by Western an-alysis (fig. S12). By using triple quadrupole massspectrometry, we could detect 5caC in genomicDNA isolated from TDG-depleted ES cells, butno reliable signal was detected in TDG-proficientcontrol cells expressing scramble short hairpinRNA (shRNA) (Fig. 4E). Similarly, 5caC wasdetectable in mouse induced pluripotent stem(iPS) cells when the Tdg gene was knocked out(fig. S13). Judging from our calculation basedon the measurement of a 5caC standard, the num-ber of 5caC per genome is ~9000 in Tdg-depletedES or iPS cells but below 1000 in wild-type cells.

TDG has been implicated in DNA demeth-ylation for its function in excising the deami-nation product of 5mC, 5hmC, or 5mC itselffrom DNA (17–19), yet mammalian TDG lacksglycosylase activity toward 5mC (6, 12). Al-though TDG is able to excise 5hmU (19), thedeamination product of 5hmC, our work pro-vides evidence that the Tet dioxygenases oxi-dize 5mC and 5hmC to 5caC, which becomesa substrate for TDG. Therefore, Tet-mediatedconversion of 5mC and 5hmC to 5caC couldtrigger TDG-initiated BER, as indicated here.These sequential events would lead to DNAdemethylation, because unmethylated cytosinesare inserted into the repaired genomic region(fig. S14).

Genome-wide mapping revealed that Tet1 isrelatively enriched in CpG-rich active promotersthat are unmethylated (20–23), but 5hmC is un-derrepresented in the majority of Tet1 bindingsites in ES cells (24–26). These apparent para-doxes might be accounted for if active pro-moters with Tet1 binding sites were preventedfrom erroneous hypermethylation because of Tet1oxidizing 5mC into 5caC, which could then beremoved by TDG-mediated BER repair. In thiscase, 5mC is most likely undetectable in the ac-tive promoters because of their transient exis-tence in a small proportion of cells. Likewise, inmany of the Tet1 binding sites, 5hmC could beunderrepresented because of conversion to 5caC,which is rapidly removed in cells.

Note added in proof: During the revisionof this manuscript, Ito et al.’s report (www.sciencemag.org/content/early/2011/07/20/science.1210597.abstract) appeared online de-scribing the enzymatic activity of Tet proteinsin the conversion of 5mC to 5fC and 5caC, aswell as the detection of these derivatives in mousegenomic DNA.

References and Notes1. R. Jaenisch, A. Bird, Nat. Genet. 33 (suppl.), 245

(2003).2. S. Simonsson, J. Gurdon, Nat. Cell Biol. 6, 984

(2004).3. X. J. He, T. Chen, J. K. Zhu, Cell Res. 21, 442

(2011).4. Z. Liutkeviciute, G. Lukinavicius, V. Masevicius,

D. Daujotyte, S. Klimasauskas, Nat. Chem. Biol. 5, 400(2009).

5. C. P. Walsh, G. L. Xu, Curr. Top. Microbiol. Immunol.301, 283 (2006).

6. S. C. Wu, Y. Zhang, Nat. Rev. Mol. Cell Biol. 11, 607(2010).

7. C. Dahl, K. Grønbæk, P. Guldberg, Clin. Chim. Acta 412,831 (2011).

8. M. Tahiliani et al., Science 324, 930 (2009).9. T. Pfaffeneder et al., Angew. Chem. Int. Ed. Engl. 50,

7008 (2011).10. D. Globisch et al., PLoS ONE 5, e15367 (2010).11. T. Lindahl, R. D. Wood, Science 286, 1897 (1999).12. D. Cortázar et al., Nature 470, 419 (2011).13. M. T. Bennett et al., J. Am. Chem. Soc. 128, 12510

(2006).14. B. Hendrich, U. Hardeland, H. H. Ng, J. Jiricny, A. Bird,

Nature 401, 301 (1999).15. H. E. Krokan, R. Standal, G. Slupphaug, Biochem. J. 325,

1 (1997).16. R. J. Boorstein et al., J. Biol. Chem. 276, 41991

(2001).17. R. Métivier et al., Nature 452, 45 (2008).18. B. Zhu et al., Proc. Natl. Acad. Sci. U.S.A. 97, 5135

(2000).19. S. Cortellino et al., Cell 146, 67 (2011).20. W. A. Pastor et al., Nature 473, 394 (2011).21. G. Ficz et al., Nature 473, 398 (2011).22. C. X. Song et al., Nat. Biotechnol. 29, 68 (2011).23. H. Wu et al., Genes Dev. 25, 679 (2011).24. K. Williams et al., Nature 473, 343 (2011).25. Y. Xu et al., Mol. Cell 42, 451 (2011).26. H. Wu et al., Nature 473, 389 (2011).Acknowledgments: We thank C. Walsh for critical reading

of the manuscript, G. Shi and S. Klimasauskas fordiscussions, J. Ju for providing Tet cDNA clones, T. Carellfor 2′-deoxy-5-carboxylcytidine and Z. Hua for the TDGantibody. This study was supported by grants from theMinistry of Science and Technology China (2007CB947503and 2009CB941101 to G.-L.X., 2010CB912100 to L.L.),National Science Foundation of China (30730059 toG.-L.X., 30930052 and 30821065 to L.L.), and theStrategic Priority Research Program of the Chinese Academyof Sciences (XDA01010301 to G.-L.X.) and by the NIH(GM071440 to C.H.) and (1S10RR027643-01 to K.Z.).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1210944/DC1Materials and MethodsFigs. S1 to S14References (27–32)

11 July 2011; accepted 25 July 2011Published online 4 August 2011;10.1126/science.1210944

Multi-Input RNAi-Based Logic Circuitfor Identification of SpecificCancer CellsZhen Xie,1,2* Liliana Wroblewska,2 Laura Prochazka,3 Ron Weiss,2,4† Yaakov Benenson1,3†‡

Engineered biological systems that integrate multi-input sensing, sophisticated informationprocessing, and precisely regulated actuation in living cells could be useful in a variety ofapplications. For example, anticancer therapies could be engineered to detect and respond tocomplex cellular conditions in individual cells with high specificity. Here, we show a scalabletranscriptional/posttranscriptional synthetic regulatory circuit—a cell-type “classifier”—that sensesexpression levels of a customizable set of endogenous microRNAs and triggers a cellular responseonly if the expression levels match a predetermined profile of interest. We demonstrate that a HeLacancer cell classifier selectively identifies HeLa cells and triggers apoptosis without affectingnon-HeLa cell types. This approach also provides a general platform for programmed responsesto other complex cell states.

Synthetic biomolecular pathways with elab-orate information processing capabilitywill enable in situ response to complex

physiological conditions (1). For example, mul-

tiple cancer-specific biomarkers (2) can be sensedand integrated in such pathways, providing pre-cise control of therapeutic agents (3). A combina-tion of up to two tissue-specific signals, including

promoter and/or microRNA (miRNA) activity,mRNA, and protein levels, have been used topartially restrict therapeutic action to cancer cells(4–9). In parallel, research in synthetic biology hasdemonstrated multi-input information processingin living cells (10–18), but the interaction be-tween these systems and the cellular context hasbeen limited. Yet, general-purpose mechanismsfor programmable integration of multiple mark-ers are required to detect cell state and precisely

1Faculty of Arts and Sciences (FAS) Center for SystemsBiology, Harvard University, 52 Oxford Street, Cambridge,MA 02138, USA. 2Department of Biological Engineering,Massachusetts Institute of Technology (MIT), 40 Ames Street,Cambridge, MA 02142, USA. 3Department of Biosystems Sci-ence and Engineering, Eidgenössische Technische Hochschule(ETH) Zürich, Mattenstrasse 26, Basel 4058, Switzerland. 4De-partment of Electrical Engineering and Computer Science, MIT,40 Ames Street, Cambridge MA 02142, USA.

*Present address: Department of Biological Engineering,MIT, 40 Ames Street, Cambridge, MA 02142, USA.†To whom correspondence should be addressed. E-mail:[email protected] (Y.B.); [email protected] (R.W.)‡Present address: Department of Biosystems Science andEngineering, ETH Zürich, Mattenstrasse 26, Basel 4058,Switzerland.

www.sciencemag.org SCIENCE VOL 333 2 SEPTEMBER 2011 1307

REPORTS

Biocomputer

!12

regulate therapeutic actuation. Here, we describesuch a mechanism, a “classifier” gene circuit thatintegrates sensory information from a large numberof molecular markers to determine whether a cellis in a specific state and, if so, produces a biolog-ically active protein output. Specifically, whentransiently expressed inside a cell our classifierascertains whether the expression profile of sixendogenousmiRNAs (19)matches a predeterminedreference profile characteristic of the HeLa cer-vical cancer cell line. A match identifies the cellas HeLa and triggers apoptosis (Fig. 1A).

A HeLa reference profile was constructedby choosing a set of HeLa-high and HeLa-lowmarkers so that HeLa cells express high or lowlevels of these, respectively, whereas expressionin any healthy cell differs substantially fromthat of a typical HeLa cell for at least one ofthese markers. Multiple reference profiles arepossible, and these profiles do not need to dif-ferentiate HeLa from other cancers. We designeda sensor motif for HeLa-high markers compris-ing a “double-inversion” module that allowsoutput expression only if the marker is presentat or above its level in HeLa cells but efficientlyrepresses the output if the marker’s level is low.The design is based on our previously describedmodule that consists of the small interfering RNA(siRNA)–targeted transcriptional Lac repressor(LacI) and a LacI-controlled promoter CAGop(CAG promoter followed by an intron with twoLacO sites) (17). We improved the ON:OFF ratioof our original design to approximately 8- to

10-fold by introducing a reverse tetracycline-controlled transactivator (rtTA) to control LacIand by targeting both the repressor and the acti-vator using miRNA in a feed-forward loop (Fig.1B and fig. S1) (20, 21). A HeLa-low markersensor was implemented by fusing four repeats(5) of fully complementary target sites into theoutput’s 3′-untranslated region (3′-UTR) (Fig.1C). A complete classifier circuit consists of aset of HeLa-high and HeLa-lowmarker sensorsall arranged to target the same output (Fig. 1D).Multiple HeLa-low sensors are combined byfusing their corresponding miRNA targets inthe output’s 3′-UTR (17). Multiple HeLa-highsensors are integrated by regulating the sameoutput in parallel. Accordingly, all HeLa-highmarkers must be present at the same time toenable output expression from CAGop becauseany individual double-inversion module can ef-ficiently repress the output by itself (22).

To choose HeLa markers for the referenceprofile and analyze expected circuit performance,we created a mathematical model consisting ofa multi-variable circuit response function that usesexperimentally derived responses (23) of individ-ual sensors to their miRNA inputs [fig. S2 andsupporting online material (SOM) text]. We firstevaluated HeLa-high marker combinations fromthe set of the top 12 candidates determined byanalyzing expression data from the MicroRNAAtlas (fig. S3) (24). On the basis of reasonableassumptions regarding sensor response param-eters (22), we found that using miR-21 togeth-

er with a composite marker that includes bothmiR-17 and miR-30a (miR-17-30a) results inat least a fivefold difference between circuit out-put in HeLa cells and the output in all but a fewhealthy cell types profiled in the MicroRNAAtlas(Fig. 2A and fig. S2D) (24). We then searchedfor HeLa-lowmarkers and found that miR-141,miR-142(3p), andmiR-146a are highly expressedin the potentially misclassified cell types but un-expressed in HeLa (Fig. 2B). With inclusion ofthese inputs in the response function, the modelpredicts at least a sixfold output difference inHeLa cells relative to the closest other cell typeUSSC-7d and on average about a 160-fold dif-ference relative to the rest of the cells (fig. S2E).We also computed selectivity of all possiblemarkersubsets and found that it steadily increases asthe number of markers goes from one to five(Fig. 2C). With all inputs included, the responsefunction is well approximated by a Boolean ex-pression (fig. S2F):

miR‐21 AND miR‐17‐30aAND NOT(miR‐141)

AND NOT[miR‐142(3p)]AND NOT(miR‐146a)

In addition to HeLa, we tested circuit operationin six control cell lines. For practical reasons,most of these are cancer cell lines that can behandled with ease and that are predicted by themodel to produce less than 1% of HeLa output(22, 24). First, we measured all miRNA activities

Fig. 1. High-level architecture of a cell typeclassifier. (A) Schematic representation of aHeLa-specific classifier circuit operation. Graycircle, healthy cells; light green, HeLa cells.(B) High-level and detailed description ofdouble-inversion module for sensing HeLa-high miRNAs. Act, activator; R, repressor. (C)High-level and detailed description of HeLa-low miRNA sensor. (D) Schematic representa-tion of an integrated multi-input classifier.Lines with bars indicate down-regulation. R1and R2 represent double-inversion modules.The entire network implements a multi-inputAND-like logic function for identification andselective killing of HeLa cells through regu-lated expression of hBax.

miR-X

Output

B

DC

Output mRNA

miR-X

DNAOutput

rtTA

miR-X

LacI

rtTA

LacI Output

CMV

pTRE

CAG LacO2

CAGop

hBax Protein

hBaxCla

ssifi

er c

ircui

t

HeLa-low-1

R1

HeLa-high-1 HeLa-high-2

R2

HeLa-low-2 HeLa-low-3

Apoptosis

A

Profile matching

No match Match

No effect Apoptosis

HeLa classifier circuit

Normal cell

HeLacell

Act

R

OutputD

oubl

e-in

vers

ion

mod

ule

miR-X

2 SEPTEMBER 2011 VOL 333 SCIENCE www.sciencemag.org1308

REPORTS

Biocomputer

!12










miR-X

Output

B

DC

Output mRNA

miR-X

DNAOutput

rtTA

miR-X

LacI

rtTA

LacI Output

CMV

pTRE

CAG LacO2

CAGop

hBax Protein

hBaxCla

ssifi

er c

ircui

t

HeLa-low-1

R1


R2


Apoptosis

A

Profile matching

No match Match

No effect Apoptosis


Normal cell

HeLacell

Act

R

OutputD

oubl

e-in

vers

ion

mod

ule

miR-X


REPORTS










miR-X

Output

B

DC

Output mRNA

miR-X

DNAOutput

rtTA

miR-X

LacI

rtTA

LacI Output

CMV

pTRE

CAG LacO2

CAGop

hBax Protein

hBaxCla

ssifi

er c

ircui

t

HeLa-low-1

R1


R2


Apoptosis

A

Profile matching

No match Match

No effect Apoptosis


Normal cell

HeLacell

Act

R

Output

Dou

ble-

inve

rsio

n m

odul

e

miR-X


REPORTS

Cyborg Yeast

!13

Cyborg Yeast• Feedback control is a central idea in engineering.

!13


• It involves applying an action on a system based on the error between its actual output and a desired reference value.

!13



!13



!13

Temperature



!13

Temperature21˚C



!13

Temperature21˚C+/-

Error



!13

TemperatureThermostat

Heating on/off21˚C

+/-

Error



!13


Heating on/off21˚C

+/-

Error



!13


Heating on/off21˚C

+/-

Error

Cyborg Yeast• Can feedback control also be applied to gene

expression to achieve the robust regulation of a protein of interest?

!14

NATURE BIOTECHNOLOGY VOLUME 29 NUMBER 12 DECEMBER 2011 1115

B R I E F C O M M U N I C AT I O N S

measurements, the system model and knowledge of the pulse history— to generate an estimate of the unmeasured states of the gene expression model. Based on this estimate, the feedback algorithm then calculates, in real time, the train of light pulses that will minimize the deviation of the model-predicted output from the desired fluorescence set point. Following the principle of model predictive control12, only the first pulse of this pulse train is applied to the system, and the process is repeated. This scheme successfully achieved the desired regulation levels as evidenced by the robust seven- and fourfold induction (Fig. 2b,c and Supplementary Fig. 4). Thus, in contrast to the open-loop strategy, in silico closed-loop feedback robustly achieved and maintained the desired set points despite modeling errors and biological fluctuations.

To further demonstrate the robustness of the in silico feedback design, we perturbed the system with a train of red and far-red pulses over a period of 3 h to drive YFP expression to an unknown initial condition. We then activated the in silico feedback controller with the task of regu-lating YFP fluorescence to fivefold above basal over a 7-h period; in all cases, regulation was robustly achieved (Fig. 2d and Supplementary Fig. 4). Notably, in this case, an open-loop train of light pulses, even one based on a high-fidelity model of the process13, could not possibly be used because the initial state of the system was unknown.

Until feedback control of biological systems can be genetically encoded in vivo to behave deterministically and without cross-talk,

biology is far from being a predictable engineering medium like electronics. But by interfacing electronic control with biological responses, in silico feedback provides an approach for unprec-edented, quantitative control over the activity of living cells. This technology should find wide application in systems and synthetic biology. Probing endogenous biological circuits often yields unpre-dictable results because of compensatory cellular regulation that alters the expression in dynamic and complex ways. In such scenarios, an in silico feedback module could be used to ensure that protein expres-sion remains tightly regulated at the desired level, independent of other intervening processes. The magnitude of the in silico feedback required to maintain a constant protein level could be used as a sur-rogate for the endogenous feedback present in the pathway and as a measure of its homeostatic capabilities. The ability to exert precise in silico control of engineered biological circuits based on direct readings of intracellular states will also facilitate the use of synthetic systems for biotechnological applications. For example, in the pro-duction of biofuels or small-molecule drugs, an in silico feedback module could be deployed to regulate the levels of toxic by-products that invariably ensue from manipulation of metabolic pathways. Finally, in silico feedback offers exciting therapeutic opportunities. As real-time physiological readouts become available, one can envi-sion the possibility of using real-time closed-loop control to achieve more regulated and precise interventions.

Figure 1 Characterization of the light-switched system. (a) Light-switchable gene system based on PhyB-PIF3 interaction. Transformed cells grown in darkness and incubated with the chromophore phycocyanobilin (PCB) synthesize both PhyB(Pr)-GBD and PIF3-GAD fusion proteins. Because PIF3 interacts only with the activated form of PhyB (Pfr), the Gal1 target gene is initially off. Upon exposure to red light, PhyB is rapidly converted into its active Pfr form and binds the PIF3 moiety of PIF3-GAD. The transcription activation domain of Gal4 is therefore recruited to the promoter and induces transcription of the target gene. Exposure to far-red light switches off gene expression by rapidly converting PhyB into its inactive Pr form, causing its dissociation from PIF3-GAD. (b) The expression of YFP driven by the Gal1 promoter can be repeatedly switched on and off using a train of red (R) and far-red (FR) pulses. The trains of light pulses can serve as a control input, and the amount of YFP plays the role of a controlled output. (c) Experimental dynamics of cell fluorescence in response to a red pulse followed by a far-red pulse. All pulses have 1-min duration. YFP flow cytometry measurements (squares) were taken every 30 min. Each set of matched colored arrow and output squares and curve represent a distinct experiment in which the far-red input was applied at different times. Spontaneous transition of PhyB Pfr to Pr takes place in the dark (a phenomenon known as ‘dark reversion’) resulting in dissociation of PIF3 from PhyB. Consequently, cell fluorescence reaches a peak and then declines, as mRNA decays over time. Gray squares and line correspond to a control experiment with chromophore addition and no light exposure. (d) Simulation of the response to the same input as in c. The model reproduces several essential features of the experimental responses, including peak times and decay dynamics. Slight differences between simulated and experimental responses are due to nonlinear effects and delays that are not captured by the model. (e) Reversibility of the PhyB-PIF3 interaction. The system does not lose its responsiveness to light over several on-off cycles. (f) Simulation results for the same input as in e. (g) Response to multiple red pulses. Multiple applications of red light drive the system to higher expression levels than a single red pulse. (h) Simulation results for multiple- compared with single-pulse responses.

Gal4 AD

a c d

fe

g hb

PIF3

PhyBPr

Gal4 BDGal1 UAS

Gal4 ADPIF3

PhyB PfrGal4 BDGal1 UAS

R FR FR YFP signalR

Red

ligh

t(6

50 n

m)

Far-red light(730 nm

)

Venus

Venus

765

5

4

3

2

18

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

14

10

6

2

1

4321

FR

Experiment Simulation

FR

FR

FR

FRFR

FR

FR

ControlR

R

RR

RR

R

RR

RR

R

R

R

0 50 100 150 200 250 300 350

0 50 100 150 200 250 300

0 50 100 150 200 250 300Time (min) Time (min)

Time (min) Time (min)


18

14

10

6

20 50 100 150 200 250 300

5

4

3

2

1

FRFR

FR

FR

R

R

R0 50 100 150 200 250 300

7654321

FRFR

FR

FRR

0 50 100 150 200 250 300 350

Control

1114 VOLUME 29 NUMBER 12 DECEMBER 2011 NATURE BIOTECHNOLOGY


Regulating a dynamic system using feedback control involves process-ing measurements of its output in real time to determine appropri-ate inputs designed to drive its behavior to follow a desired pattern. Because feedback control enables robust regulation in the face of uncertainty and disturbances, it is a recurring theme at every level of organization throughout biology and engineering. Although evolved biological feedback mechanisms found in nature appear robust, engi-neering feedback control schemes in cells to produce new functions has proven to be a tedious, iterative process1–6 that has been achieved with limited success. Here we propose the concept of in silico feedback control as a complement to feedback circuitry built from biological components. In silico feedback uses computational control algorithms running on a digital computer and updated with real-time measure-ment data. The algorithms prescribe external inputs that achieve and maintain a desired circuit behavior while automatically compensating for circuit variability.

The implementation of in silico feedback requires an external input and biological modules that can respond to that input. Working with S. cerevisiae, we took advantage of a system that has been used to control diverse biological processes7–9, the light-responsive Phy/PIF module10 (Fig. 1a,b). Upon ligation to the small-molecule chromo-phore phycocyanobilin (PCB), the plant photoreceptor chromo-protein PhyB undergoes a light-gated interaction with phytochrome interacting factor (PIF). Two fusion constructs—the photosensory domain of PhyB fused to the Gal4 DNA-binding domain (PhyB-GBD)

In silico feedback for in vivo regulation of a gene expression circuitAndreas Milias-Argeitis1,4, Sean Summers1,4, Jacob Stewart-Ornstein2,4, Ignacio Zuleta2, David Pincus2, Hana El-Samad2, Mustafa Khammash3 & John Lygeros1

We show that difficulties in regulating cellular behavior with synthetic biological circuits may be circumvented using in silico feedback control. By tracking a circuit’s output in Saccharomyces cerevisiae in real time, we precisely control its behavior using an in silico feedback algorithm to compute regulatory inputs implemented through a genetically encoded light-responsive module. Moving control functions outside the cell should enable more sophisticated manipulation of cellular processes whenever real-time measurements of cellular variables are possible.

and PIF3 fused to the Gal4 activation domain (PIF3-GAD)—allow one to use red (650 nm) and far-red (730 nm) pulses of light to switch on and off, respectively, the transcription of Gal4-responsive genes in S. cerevisiae10. In particular, we used cells containing a YFP reporter driven by the Gal1 promoter, which contains Gal4 binding sites.

To probe the dynamic behavior of the system and to devise a computational model for in silico feedback control, we modeled the dynamics of the Phy/PIF/Gal system using a simple fourth-order linear ordinary differential equation. We then performed several time-course ‘identification experiments’ to excite the crucial responses of the system and estimate the model’s five free parameters (Fig. 1c–h and Supplementary Methods). These experiments involved stimulat-ing cells with different trains of red and far-red light pulses delivered using a custom-built ‘light pulser’ (Supplementary Fig. 1) and then obtaining single-cell fluorescence measurements by flow cytometry. In each experiment, we reproducibly initialized the system at its ‘station-ary condition’, characterized by constant growth rate and basal fluo-rescence level, and we computed the fold change as induction of the system above this initial basal fluorescence (Supplementary Figs. 2 and 3). Despite the intrinsic ability of the system to achieve rapid acti-vation and shutoff, the slow maturation dynamics and long half-life of the YFP reporter protein generate a delayed and time-integrated snapshot of the overall system dynamics, making control of the system particularly challenging. The mathematical model with optimized parameters captured the essential features of the data (Fig. 1d,f,h).

Using the model as a starting point, we investigated strategies to robustly regulate in vivo the average YFP fluorescence of the Phy/PIF/Gal system to desired levels, or ‘set points’, in this case seven- and four-fold above basal levels over a 7-h period (Fig. 2). First nonfeedback strategies (also known as open-loop control) were investigated. One such strategy is to select a control signal that, according to the model, will achieve the desired objective and then to apply this signal as an input to the process. Following this strategy, we computed trains of light pulses using the data-calibrated model and found that in com-puter simulations these pulses achieved accurate regulation of YFP (Fig. 2b,c). Yet in contrast to the simulation results, when these same trains of light pulses were experimentally administered to cells, they failed to regulate YFP intensity to the desired set points. This failure is the result of inaccuracies of the model and inevitable intracellular fluctuations.

An alternative strategy is to exploit the real-time fluorescence mea-surements to compute the control signal online, as the process evolves. To investigate this in silico closed-loop feedback control strategy (Fig. 2a), we used YFP signal measurements obtained every 30 min to compute red or far-red pulses to be applied every 15 min. Briefly, we first used a Kalman filter11—which uses information from the most recent YFP

1Department of Electrical Engineering, ETH Zurich, Automatic Control Laboratory, Zurich, Switzerland. 2Department of Biochemistry and Biophysics, University of California San Francisco, San Francisco, California, USA. 3Department of Biosystems Science and Engineering, ETH Zurich, Basel, Switzerland. 4These authors contributed equally to this work. Correspondence should be addressed to H.E.-S. ([email protected]) or M.K. ([email protected]) or J.L. ([email protected]).

Received 1 July; accepted 27 September; published online 6 November 2011; doi:10.1038/nbt.2018

Cyborg Yeast• Using a light-inducible circuit, gene expression can

be turned on or off using light.

!15

NATURE BIOTECHNOLOGY VOLUME 29 NUMBER 12 DECEMBER 2011 1115


measurements, the system model and knowledge of the pulse history— to generate an estimate of the unmeasured states of the gene expression model. Based on this estimate, the feedback algorithm then calculates, in real time, the train of light pulses that will minimize the deviation of the model-predicted output from the desired fluorescence set point. Following the principle of model predictive control12, only the first pulse of this pulse train is applied to the system, and the process is repeated. This scheme successfully achieved the desired regulation levels as evidenced by the robust seven- and fourfold induction (Fig. 2b,c and Supplementary Fig. 4). Thus, in contrast to the open-loop strategy, in silico closed-loop feedback robustly achieved and maintained the desired set points despite modeling errors and biological fluctuations.

To further demonstrate the robustness of the in silico feedback design, we perturbed the system with a train of red and far-red pulses over a period of 3 h to drive YFP expression to an unknown initial condition. We then activated the in silico feedback controller with the task of regu-lating YFP fluorescence to fivefold above basal over a 7-h period; in all cases, regulation was robustly achieved (Fig. 2d and Supplementary Fig. 4). Notably, in this case, an open-loop train of light pulses, even one based on a high-fidelity model of the process13, could not possibly be used because the initial state of the system was unknown.

Until feedback control of biological systems can be genetically encoded in vivo to behave deterministically and without cross-talk,

biology is far from being a predictable engineering medium like electronics. But by interfacing electronic control with biological responses, in silico feedback provides an approach for unprec-edented, quantitative control over the activity of living cells. This technology should find wide application in systems and synthetic biology. Probing endogenous biological circuits often yields unpre-dictable results because of compensatory cellular regulation that alters the expression in dynamic and complex ways. In such scenarios, an in silico feedback module could be used to ensure that protein expres-sion remains tightly regulated at the desired level, independent of other intervening processes. The magnitude of the in silico feedback required to maintain a constant protein level could be used as a sur-rogate for the endogenous feedback present in the pathway and as a measure of its homeostatic capabilities. The ability to exert precise in silico control of engineered biological circuits based on direct readings of intracellular states will also facilitate the use of synthetic systems for biotechnological applications. For example, in the pro-duction of biofuels or small-molecule drugs, an in silico feedback module could be deployed to regulate the levels of toxic by-products that invariably ensue from manipulation of metabolic pathways. Finally, in silico feedback offers exciting therapeutic opportunities. As real-time physiological readouts become available, one can envi-sion the possibility of using real-time closed-loop control to achieve more regulated and precise interventions.

Figure 1 Characterization of the light-switched system. (a) Light-switchable gene system based on PhyB-PIF3 interaction. Transformed cells grown in darkness and incubated with the chromophore phycocyanobilin (PCB) synthesize both PhyB(Pr)-GBD and PIF3-GAD fusion proteins. Because PIF3 interacts only with the activated form of PhyB (Pfr), the Gal1 target gene is initially off. Upon exposure to red light, PhyB is rapidly converted into its active Pfr form and binds the PIF3 moiety of PIF3-GAD. The transcription activation domain of Gal4 is therefore recruited to the promoter and induces transcription of the target gene. Exposure to far-red light switches off gene expression by rapidly converting PhyB into its inactive Pr form, causing its dissociation from PIF3-GAD. (b) The expression of YFP driven by the Gal1 promoter can be repeatedly switched on and off using a train of red (R) and far-red (FR) pulses. The trains of light pulses can serve as a control input, and the amount of YFP plays the role of a controlled output. (c) Experimental dynamics of cell fluorescence in response to a red pulse followed by a far-red pulse. All pulses have 1-min duration. YFP flow cytometry measurements (squares) were taken every 30 min. Each set of matched colored arrow and output squares and curve represent a distinct experiment in which the far-red input was applied at different times. Spontaneous transition of PhyB Pfr to Pr takes place in the dark (a phenomenon known as ‘dark reversion’) resulting in dissociation of PIF3 from PhyB. Consequently, cell fluorescence reaches a peak and then declines, as mRNA decays over time. Gray squares and line correspond to a control experiment with chromophore addition and no light exposure. (d) Simulation of the response to the same input as in c. The model reproduces several essential features of the experimental responses, including peak times and decay dynamics. Slight differences between simulated and experimental responses are due to nonlinear effects and delays that are not captured by the model. (e) Reversibility of the PhyB-PIF3 interaction. The system does not lose its responsiveness to light over several on-off cycles. (f) Simulation results for the same input as in e. (g) Response to multiple red pulses. Multiple applications of red light drive the system to higher expression levels than a single red pulse. (h) Simulation results for multiple- compared with single-pulse responses.

Gal4 AD

a c d

fe

g hb

PIF3

PhyBPr

Gal4 BDGal1 UAS

Gal4 ADPIF3

PhyB PfrGal4 BDGal1 UAS

R FR FR YFP signalR

Red

ligh

t(6

50 n

m)

Far-red light(730 nm

)

Venus

Venus

765

5

4

3

2

18

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

14

10

6

2

1

4321

FR

Experiment Simulation

FR

FR

FR

FRFR

FR

FR

ControlR

R

RR

RR

R

RR

RR

R

R

R

0 50 100 150 200 250 300 350

0 50 100 150 200 250 300

0 50 100 150 200 250 300Time (min) Time (min)



18

14

10

6

20 50 100 150 200 250 300

5

4

3

2

1

FRFR

FR

FR

R

R

R0 50 100 150 200 250 300

7654321

FRFR

FR

FRR

0 50 100 150 200 250 300 350

Control

Cyborg Yeast• A computer calculates the correct sequence of

light pulses to keep the output protein at the desired level.

!16



Note: Supplementary information is available on the Nature Biotechnology website.

ACKNOWLEDGMENTSWe would like to thank P. Quail (UC, Berkeley) for the generous gift of the PIF3 construct and W. Lim (UCSF) and J. Stelling (ETH, Zurich) for providing PCB. The work was supported by National Science Foundation grant CCF-0943385 (H.E.-S.), ECCS-0835847 (M.K.), and MoVeS FP7-ICT-2009-257005 (J.L.).

AUTHORS CONTRIBUTIONSM.K. originated the concept. A.M.-A., S.S., J.S.-O., H.E.-S., J.L. and M.K. carried out the project design. A.M.-A. and S.S. did the control design. J.S.-O. did the cloning and strain construction. S.S., A.M.-A. and I.Z. designed and built the electronics. S.S. and A.M.-A. performed the experiments with J.S.-O. and with help from D.P.; H.E.-S., M.K. and J.L. supervised the project; all authors contributed to the writing of the manuscript. Corresponding authors are listed in the affiliations in alphabetical order.

COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.

Published online at http://www.nature.com/nbt/index.html. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html.

1. Lim, W.A. Nat. Rev. Mol. Cell Biol. 11, 393–403 (2010).2. Andrianantoandro, E., Basu, S., Karig, D.K. & Weiss, R. Mol. Syst. Biol. 2

2006.0028 (2006).3. Benner, S.A. & Sismour, A.M. Nat. Rev. Genet. 6, 533–543 (2005).4. Khalil, A.S. & Collins, J.J. Nat. Rev. Genet. 11, 367–379 (2010).5. Purnick, P.E. & Weiss, R. Nat. Rev. Mol. Cell Biol. 10, 410–422 (2009).6. Haynes, K.A. & Silver, P.A. J. Cell Biol. 187, 589–596 (2009).7. Levskaya, A., Weiner, O.D., Lim, W.A. & Voigt, C.A. Nature 461, 997–1001 (2009).8. Tyszkiewicz, A.B. & Muir, T.W. Nat. Methods 5, 303–305 (2008).9. Leung, D.W., Otomo, C., Chory, J. & Rosen, M.K. Proc. Natl. Acad. Sci. USA 105,

12797–12802 (2008).10. Shimizu-Sato, S., Huq, E., Tepperman, J.M. & Quail, P.H. Nat. Biotechnol. 20,

1041–1044 (2002).11. Kalman, R.E. J. Basic Eng. 82, 35–45 (1960).12. Morari, M. Comput. Chem. Eng. 23, 667–682 (1999).13. Sorokina, O. et al. J. Biol. Eng. 3, 15 (2009).

Figure 2 In silico feedback achieves robust regulation of gene expression fold change. (a) In silico feedback control scheme for the light-activated gene system. (b) Regulation of average YFP fluorescence to sevenfold over a 7-h period using in silico feedback (orange). A pre-computed light pulse train that achieves set point regulation when applied to the mathematical model (gray) did not achieve the desired fold induction when applied in open loop to the biological construct (green). In contrast, closed-loop feedback control achieves the desired fold induction. OL and CL denotes open- and closed-loop control, respectively. (c) Regulation of average YFP fluorescence to fourfold above basal over a 7-h period. Open- and closed-loop pulse trains determined as in b. (d) Regulation of average YFP to fivefold above basal over a 7-h period, starting from a randomly perturbed culture. Closed-loop control achieves the desired set point, irrespective of the initial conditions of the system.

Desiredfluorescence

a

b c d

Light pulsetrain

YFP fluorescence distribution

Kalman filter + model predictivecontroller

Kept in dark vessel with electronicallycontrolled light source

In silico feedback Cell population Output measurement

Flow cytometry

Cell culturesamples

(set point)

74

1098765432

3

2

1

6

5

4

3

2

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

1

0 50 100 150 200 250 300 350 400 0 50 100 150 200Time (min)Time (min)

Time (min)

250 300 350 400

0 50 100 150 200 250 300 350 400

OL

CL

OL CL1

CL2

CL3

CL4

CL

Cyborg Yeast

!17













Desiredfluorescence

a

b c d

Light pulsetrain





Flow cytometry

Cell culturesamples

(set point)

74

1098765432

3

2

1

6

5

4

3

2

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

1

0 50 100 150 200 250 300 350 400 0 50 100 150 200Time (min)Time (min)

Time (min)

250 300 350 400

0 50 100 150 200 250 300 350 400

OL

CL

OL CL1

CL2

CL3

CL4

CL

Cyborg Yeast

!17













Desiredfluorescence

a

b c d

Light pulsetrain





Flow cytometry

Cell culturesamples

(set point)

74

1098765432

3

2

1

6

5

4

3

2

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

Ave

rage

fluo

resc

ence

fold

cha

nge

1

0 50 100 150 200 250 300 350 400 0 50 100 150 200Time (min)Time (min)

Time (min)

250 300 350 400

0 50 100 150 200 250 300 350 400

OL

CL

OL CL1

CL2

CL3

CL4

CL

Summary

!18

Summary• Synthetic Biology is an advanced form of genetic

engineering, which aims to design artificial gene networks for a variety of exciting applications.

!18



• These include synthesis of drugs, gene therapy, biotechnology, biofuel production, sensing and removal of pollutants, etc.

!18



• These include synthesis of drugs, gene therapy, biotechnology, biofuel production, sensing and removal of pollutants, etc.

• The future of the field holds great promise, but also great challenges.

!18

Challenges

!19

Challenges• Lack of standards.

!19


• Insufficient knowledge of the underlying biology.

!19


• Insufficient knowledge of the underlying biology.

• Lack of systematic design principles and tools.

!19

Can Synthetic Biology deliver its promises?

!20

Thank you!!

Gabriele Lillacci, Ph.D.Postdoc in Control Theory and Systems Biology

Mattenstrasse 264058 BaselSwitzerland

+41 61 387 33 [email protected]/ctsb

Gabriele Lillacci, Ph.D.Postdoc in Control Theory and Systems Biology

Mattenstrasse 264058 BaselSwitzerland

+41 61 387 33 [email protected]/ctsb

Control Theory and Systems Biology Lab

PI: Prof. Dr. Mustafa Khammash

Documents

Synthetic Biology - WordPress.com · Synthetic biology is the engineering of biology: the deliberate (re)design and construction of novel biological and biologically based parts,