Building Biological Memory by Linking Positive Feedback Loops
Author(s): Dong-Eun Chang, Shelly Leung, Mariette R. Atkinson, Aaron Reifler, Daniel Forger,
Alexander J. Ninfa and Clyde A. Hutchison III
Source: Proceedings of the National Academy of Sciences of the United States of America, Vol.
107, No. 1 (Jan. 5, 2010), pp. 175-180
Published by: National Academy of Sciences
Stable URL: http://www.jstor.org/stable/40536247
Accessed: 28-11-2015 12:19 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/40536247?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

National Academy of Sciences is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the
National Academy of Sciences of the United States of America.

http://www.jstor.org

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

positive
Building
biologicalmemory
bylinking
feedbackloops
Dong-Eun Chang8, Shelly Leung3, Mariette R. Atkinson3,Aaron Reifler3,Daniel Forgerb,and Alexander J. Ninfa3'1
aDepartmentof Biological Chemistry,Universityof Michigan Medical School; and bDepartmentof Mathematics and Center for Computational Medicine and
Biology,Universityof Michigan,Ann Arbor, Ml 48109
Edited by Clyde A. HutchisonIII,The J.Craig Venter Institute,San Diego, CA, and approved November 20, 2009 (received for review July23, 2009)

ofactivator
totheconc; entration
protein
promoter
transcriptional
thanone(8, 10a kinetic
orderorsensitivity
mustdisplay
greater
are met,bistability
requirements
maybe
12). If theseminimal
conditions.
to occurundersomeenvironmental
anticipated
of
ofa bistable
Animportant
consisting
genetic
system
example
a singlepositivefeedbackloop was providedby theworkof
efwhostudiedtheso-called"preinduction
NovickandWeiner,
coli(13). Thelac opfect"ofthelacZYAoperoninEscherichia
whentheoperonis induced
is observed
eronpreinduction
effect
inas gratuitous
known
xenobiotic
inducers,
bynonphysiological
but
bind
to
the
the
cell
can
not
metabolized
that
are
ducers,
by
Lad repressor
proteinand causeit to releasethelac operator
thata
effect
refersto theobservation
DNA. The preinduction
wasrequired
forthe
inducer
ofgratuitous
concentration
higher
to maincellsthanwasrequired
induction
ofnaive(uninduced)
inducedstate.At certainconcellsin thefully
tainpreinduced
as themaintenance
known
centrations
ofinducer,
concentrations,
inducedcellsreandpreviously
naivecellsremained
uninduced,
mainedintheinduced
state,indefinitely
(13).Evenwhencultures
thelacZYAoperon
levelof induction,
showedan intermediate
cells(13,14).Aloffintheindividual
waseitherfully
on orfully
bistability| genetic network | syntheticbiology | ultrasensitive | hysteresis
effect
lacZYAoperonpreinduction
experithoughtheoriginal
and
with
TMG as the
a
chemostat
mentswereperformed
using
ofexpression inducer,
can
thattheeffect
information
weshowinthesupporting
displaya discontinuity
geneticsystems
Bistable
stablesteadystatesare obtained be demonstrated
cultures
and with
states,wheretwodistinct
flask-grown
usingstandard
without
thepresenceof stableintermediate
hasconsteadystates.The IPTG as theinducer
modeling
(Fig.SI). Mathematical
which
stablesteady firmed
determines
of
the
in thelacZYAoperonand
of bistability
themechanism
system
history
previous
stateis occupied.One of the important
andhighsensitivity
feedback
theroleofbothpositive
problemsin systems confirmed
and (14-17). Thissystem
is established
howgenetic
istounderstand
and
remainsthemostwell-characterized
bistability
biology
This is becausebistablegeneticswitchesplay an widely
usedexampleofcellularbistability.
regulated.
suchas cellular The laçZYAoperonpreinduction
ofcellularprocesses,
roleina variety
is due to thepositive
effect
important
cellcycle,andthe feedback
theeukaryotic
on itsown
oscillators,
ofthelacYproduct,
through
progression
permease,
galactoside
cellandtissuetypesin organisms expression
ofdifferentiated
inducer
development
(13,18).Thispermeaseallowsthegratuitous
to thehuman(1-7). to enterintothecell.Whencellslackthegalactoside
fromthetemperate
bacteriophage
ranging
permease,
a givencircuit as inthenaivestate,a highconcentration
studieshavefocusedon whether
is
inducer
ofgratuitous
Manyprevious
forsomerangeof required
to display
hasthecapacity
oftheoperon,
forinduction.
bistability
topology
However,
uponinduction
thepossi- the cells come to acquiremanymoleculesof the galactoside
conditions
environmental
(e.g.,reís.8-10).Although
thatthe permeasethatcanbringaboutfurther
isimportant,
itisalsoimportant
ofbistable
behavior
internalization
ofthein- z
bility
at whichit occursbe large ducer(positivefeedback).The presenceof permeaseprotein
conditions
rangeof environmental
control
ofbiological
toachievepractical
Here, molecules
a highintracellular
cellstomaintain
processes.
allowstheinduced
enough
oftheparameters concentration
andmanipulation
wefocusuponidentification
conevenwhentheextracellular
of theinducer,
atwhichbis- centration
conditions
therangeofenvironmental
thatcontrol
effectis
is low.As expected,the lac preinduction
tobe capableofbistability.eliminated
known
forsystems
is obtained
of thelacY gene(18). Furthermore,
tablity
uponmutation
We use themethodsof synthetic
wherethe
underconditions
is minimized
biologyto createmodelex- thepreinduction
effect
ofmultiple
to addressthefunctions
of
isdown-regulated
positive function
oftheLacY protein
systems
perimental
(19).Inhibition
feedback
loopsinbistability.
havearguedthattheminimal
studies
Theoretical
requirements
theremustbe sometype Author contributions:D.-E.C, D.F., and A.J.N.designed research; D.-E.C, S.L, A.R., and
aretwofold.
forgenetic
First,
bistability
of positivefeedback
Examplesof D.F. performed research; M.R.A. contributed new reagents/analytictools; D.-E.C, S.L,
controlling
geneexpression.
itsown D.F., and A.J.N.analyzed data; and D.-E.C, D.F., and A.J.N.wrote the paper.
drives
an
activator
are
when
feedback
protein
positive
of
linked
an
even
number
orwhen
negative
regulatoryThe authors declare no conflictof interest.
expression
This article is a PNAS Direct Submission.
the
a
blocks
as
when
such
are
expression
repressor
steps present,
orderor Freelyavailable online through the PNAS open access option.
ofitsownexpression.
ofa repressor
Second,thekinetic
mustbe 1To whom correspondence should be addressed. E-mail:aninfa@umich.edu.
element
feedback
tothepositive
ofthesystem
sensitivity
in thesimplecase wherea tran- This article contains supporting informationonline at www.pnas.org/cgi/content/full/
high(8, 10-12)..For example,
theresponse
ofthe 09083141 07/DCSupplemental.
itsownexpression,
activator
drives
scriptional

A common topology found in many bistable genetic systemsis two
interactingpositive feedback loops. Here we explore how this
relativelysimple topology can allow testabilityover a large range
of cellular conditions. On the basis of theoretical arguments, we
predictthat nonlinear interactionsbetween two positive feedback
loops can produce an ultrasensitive response that increases the
range of cellular conditions at which bistabilityis observed. This
predictionwas experimentallytested by constructinga synthetic
genetic circuitin Escherichiacoli containingtwo well-characterized
positivefeedback loops, linkedin a coherentfashion.The concerted
action of both positivefeedback loops resulted in bistable behavior
over a broad range of inducer concentrations;when either of the
feedback loops was removed, the range of inducerconcentrations
at whichthe systemexhibited bistabilitywas decreased by an order
of magnitude. Furthermore,bistabilityof the system could be
tuned by alteringgrowth conditionsthat regulate the contribution
of one of the feedback loops. Our theoretical and experimental
work shows how linked positive feedback loops may produce the
robust bistable responses required in cellular networks that regulate development,the cell cycle,and manyother cellularresponses.

UJ

www.pnas.org/cgi/doi/10. 1073/pnas.09083 141 07

PNAS

| January 5, 2010

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

| vol. 107

| no. 1 | 175-180

occurswhenthePTS componentand Results
theLacY permeaseactivity
protein EIIAglc is present in its un- A GraphicAnalysisof the Problem.To understandthe factors
signal-transduction
stateand bindsto LacY (20). This occurswhen controlling
of theactivatormoduleof the
therangeofbistability
phosphorylated
the cell is grownin the presenceof PTS sugars,such as glucose, Atkinsonet al. oscillator,we use rate-balanceplotsfollowing
the
and, to lesserextents,whenthe cell is grownin the presenceof workof Ferrelland Xiong (10, 11). In thismethod,the ratesof
othersubstratesthatexertcataboliterepression(21). Thus, the activatorproductionand decay are plottedas a functionof accells tivatorconcentration.
effectwas notdiscerniblein glucose-grown
lac preinduction
Because activatorhas no directeffecton
cells, the lac preinduction its own decay,we expectactivatordecay to be a simplelinear
(Fig. SI). Even in succinate-grown
in an experiment
effectwas a fairlyweak bistability;
usingIPTG
functionof activatorconcentration(Fig. 1). Conversely,bioas the inducer,the range of inducerconcentrationsat which chemicalstudiesoftheactivationof
transcription
bytheNRI ~ P
was observedwas narrow[about4-foldin flask-grown activatorindicateda
bistability
high kineticorder and an S-shaped reto
cells (Fig. SI)]. Indeed, in flask-grown
cells, it was difficult
sponse (25) (Fig. 1). Steadystatesare possible onlywherethe
of IPTG (Fig. SI).
demonstrate
a maintenanceconcentration
curvesforproductionrateof activatorand decayrateofactivator
In nature,simplebistablesystemswitha singlepositivefeed- intersect
(Fig. 1). A keyfeatureof thegenetictoggleswitchused
back loop are rarelyencountered;instead,naturalsystemsare in our studies
is that activatorand repressorcompete for the
complex and contain multiplefeedback loops that could be promoterthat drives transcription
of the activatorstructural
director indirectand usinga varietyofbiochemicalmechanisms,
In this geneticsystem,activatorhas no direct
gene
(Fig.
L4).
mechanisms.Fourteen effecton the Lad
in combination
withadditionalregulatory
repressor,which is presentat a low conexamplesof such systemshave been noted by Brandmanet al. stitutivelevel. The inducerIPTG servesto decrease the con(22). For example, in the genetic systemthat controlspro- centrationof functionalrepressor.Thus, changesin the IPTG
gressionthroughthe cell cycle,the mitotictriggerproteinCdc2 concentration
the curveforpromaybe thoughtof as shifting
participatesin threepositivefeedbackloops (Cdc2 -» Cdc25 -» ductionof activatorto the
the
as
or
left,
depictedin Fig. 1.
right
Cdc2; Cdc2-IWee-l-l Cdc2; Cdc2-IMytl-ICdc2) (22). Similarly, The
at whichthe systemdisplays
of IPTG concentrations
range
fortraversalof the startof the cell cyclein buddingyeast,the
is thuslimitedto the extentto whichthe production
Cdc28 protein participatesin two positive feedback loops bistability
curve
be shiftedto therightor theleftwhilestillmaintaining
may
(Cdc28-I Sicl-I Cdc28; Cdc28 -+ Cln -» Cdc28). Other systems at least two
withthe decaycurve(Fig. 1).
pointsof intersection
include
those
responshowingmultiplepositivefeedbackloops
In the systemconsideredabove,activatorproductionratedismammalian
sibleforp53 regulation,
Xenopusoocytematuration,
a highsensitivity,
whereasactivatordecayratewas linear,
calciumsignaltransduction,
eukaryoticchemotaxis,B cell fate played
in
accordance
with
the
systemwe use. However,it
experimental
EGF receptorsignaling,blood clotting,and plaspecification,
a
noted
that
nonlinear
should
be
degradationrate could
having
telet aggregation(noted in ref.22). The Bcl2 apoptoticswitch
This
result
was
also
achieve
bistability.
presentedbyFerrell
providesan additionalexample,wheretwoindependentpositive and help who considered
a
transduction
system(26).
signal
Xiong,
feedbackloops participatein producingbistability
(23).
Simpleinspectionof Fig. 1 revealstwo of the keyparameters
Foundational work on the behavior of biological systems
therangeofinducerconcentrations
at whichthesystem
affecting
containingmultiplefeedbackloops was presentedby Thomas will
The firstof these is the steepness(sensidisplaybistability.
and D'Ari (24). Thisworkshowedhow thepresenceof multiple
kinetic order) of the activatorproductioncurve; the
feedbackloops could lead to an unexpectedly
large numberof tivity,
thisresponsecurveis,thegreateritmaybe shiftedto the
steeper
states.
steady
twointersections
withthe
left
or
the
rightwhilestillmaintaining
Two additionalhypothesesforthe presenceof multipleposthe rangeof intimescalesof the loops decay curve.A second keyparameteraffecting
itivefeedbackloops are that different
at whichbistability
is observedis the abducer concentrations
(22)
provideforresistanceto noise undercertaincircumstances
solute magnitudeof the activator'seffecton itself.That is, the
robustto certainparametervariations(23).
or make bistability
Ferrellhas noted thatcoherentlinkageof a positivefeedback absoluteheightof theS-shapedactivatorproductioncurvelimits
thedistanceit maybe displacedto theleftor therightwhilestill
loop and a double-negativefeedbackloop in a systemof opwiththe decay curve.In
over a wide range of maintainingat least two intersections
posingenzymescould resultin bistability
conditions(10). In naturalsystemswithcomplexcircuitarchi- this work,we focus on controllingthe steepnessof the protectures,testingtherolesof themultiplepositivefeedbackloops ductioncurve.In supportof the simplegraphicmethodused in
and muchof theworkin thisarea to date has been Fig. 1, a more formalanalysisof our specificsystemleadingto
is nontrivial
the same conclusionsis presentedin the SI Text.
theoretical.
purely
We developedan experimental
multiloopsystembycombining
the "activatormodule" of the syntheticgenetic oscillatorof HowCanthe Steepness oftheActivatorProductionCurveBe Increased?
of a promoterto its activatoris dependentupon
Atkinsonet al. (12Jwiththegalactosidepermeasefeedbackloop The sensitivity
ofthelacZYA operon.The activatormoduleoftheAtkinsonet al. numerousfactors,such as the oligomericstate of the activator
oscillator,when placed into cells that expressLad repressor protein,the numberof molecules that are requiredfor tranformsa genetictoggleswitchin whichan activator scriptionalactivation,and the details of the interactionsof
constitutively,
the
formof theglnG product(NRI or activatorand polymerasewitheach otherand withothermacprotein, phosphorylated
oftheglnGstructural
gene,and romoleculesthatinteractwiththemand thereforecompetefor
NtrC),activatesthetranscription
to adjustin a
Lad repressestranscription
of theglnG gene. Because activator them.In mostcases, theseparametersare difficult
and repressorcompeteforcontrolof transcription
of theglnG systematic
way and thus a general solutionto the problemof
withthe level of IPTG re- increasingthe apparentkineticorderof an activator'seffecton
gene, the systemdisplayshysteresis,
of induction itselfmustemploya different
approach.It has longbeen known
quiredforinductiondependenton thepriorhistory
of thissystemwas moreprominent
thanthat thatultrasensitive
responsesto a stimulusmay be obtainedin
(12). The hysteresis
systemsconsistingof linkedcyclesof reverdisplayedby the galactosidepermeasesystem,in thatbistability signaltransduction
whenthestimulusregulatesmultiple
was observedover an «10-fold range of IPTG concentrations siblecovalentmodification,
(12). We showthatthemultiloopsystemobtainedbylinkingthe distinctactivitiesin the signalingsystem(27) (reviewedin ref.
is referredto as "multistepultraactivatormodule and galactosidepermeasefeedbackloops dis- 28). Such ultrasensitivity
could be sensitivity"
played extensivebistability,
indicatingthat bistability
(27). Recently,Rossi and colleagues used a model
builtup bylinkageof distinctfeedbackloops.
systemto showthattheapparentkineticorderof a
experimental
176

| www.pnas.org/cgi/doi/IO.IOyB/pnas.OSOSBMIO?

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

Chang et al.

Fig. 1. Design and functionof the genetic toggle switch. (A) Basic circuitdesign for the genetic toggle switch.The Lacl repressorwas produced fromthe
naturalwild-typelad gene. The activatormodule containingthe glnG structuralgene was located in the rbs region of the E. coli chromosome. The activator
and repressorcompeted forcontrolof the expressionof the activatormodule promoter;when neitheractivator nor repressorproteinswere present,a weak
promoter(not depicted) allowed for transcriptionof the activator gene (12). The reporterconsisted of a fusion of the activator-dependentglnK promoter
withthe lacZ gene, located in the trp region of the chromosome. Forthe systemas shown with a single positivefeedback loop, the lacY gene was nextto lad,
but contained a null mutation. (B) Graphicrepresentationof the factorsaffectingthe range of IPTGconcentrationsat which bistabilityis obtained. All of the
plots depict rate of activatorproductionor destructionvs. the concentrationof activator.The destructionof activatoris not regulated by activatorand thus is
likelyto have a slope of 1, as depicted. Conversely,the production of activator is known to display high kineticorder and is depicted as the S-shaped curve.
The role of IPTG is to shiftthis S-shaped curve to the rightor the left,as depicted. Steady states occur when the two curves intersect,as indicated by small
circles.The dotted circlein the Center plot depicts an unstable steady state. Factorscontrollingthe shape of the S-shaped activator productioncurve,such as
the steepness of this curve or its absolute height/controlthe range of bistabilityof the system.

responseto a stimuluswas increasedwhen the
transcriptional
stimulusaffectedboth the activationand the repressionof a
promoter,relativeto the situationswherea singlefunctionwas
regulated(29). We regardthisresultas a special class of multiand reasonedthatwe couldsimilarly
increase
stepultrasensitivity
the apparentkineticorder of activator'sresponse to itselfby
havingactivatorinfluencenotonlytheactivationof itsstructural
ungene,but also the repressionof its own gene. Furthermore,
like the systemof Rossi et al. (29), whichrequiresspecifically
engineeredproteins,we soughtto have activatorfunctionto
This was accomdecrease repressionof the systemindirectly.
plishedby havingactivatordrivethe expressionof the galactoside permease that brings about the internalizationof the
therepressor.In thebriefformalanalysis
inducerthatinactivates
of our specificsystempresentedin SI Text,we observedthatby
of repression,the
of activatorto inhibition
theactivity
extending
alreadyhighkineticorderof the responseto activatorcould be
increased.
significantly
FeedbackLoopsto Createa Composite
Well-Characterized
Combining
SystemwithCoherentLoops.The experimental
systemwithtwo
each
witha single
feedback
and
two
control
systems
loops,
positive
positivefeedbackloop,is depictedin Fig. 2. We used a previously
describedchromosomally
integratedfusionof the E. coli glnK
genes of thelacZYA operon(30). Expromoterto thestructural
pressionfromtheglnKpromoteris dependentupon theactivator
NRI ~ P. The glnKpromoter-lacZYA
ofthegenetictoggleswitch,
suchthatthenoveljointcorrespondedto
fusionwas constructed
startcodon of the lacZ gene. Thus, in the rethe translational
combinant
operon,themajorlac operonoperatorelement(lacOl)
was notpresentand neitherwas the minorlacO3 operator.The
minorlacO2 operatorwas present,as thiselementis foundwithin
intheabsenceofthemajor
thelacZ structural
gene.Nevertheless,
lac operator(lacOl), itis anticipatedthatLacl repressorwillhave
Changet al.

littledirecteffect
on transcription
ofthelacZYA structural
genesin
the recombinantcontext,whereas expressionshould be tightly
controlledbyactivator.
We chosetheglnKpromoterand thisspecificoperon fusionforour experiments
because priorworkhas
basal expression
inthe
shownthatthisoperonfusionhas negligible
absenceoftheactivator
protein(30). Aversionoi'hzglnKp-lacZYA
an internaldeletionwithinthelacY gene
operonfusioncontaining
was generatedbyrecombineering
(31). To providethemaximum
for
thegenetictoggleswitch
our
stability
synthetic
geneticsystem,
(activatormoduleof theAtkinsonet al. clock) was incorporated
intotheE. coli chromosomein therbsregion,as describedpreviously(12). The activatorofthissystemrequiresphosphorylation
foractivity,
and we providedthisfunction
byincludingwithinthe
cell an alteredkinase proteinthatbringsabout the phosphoru
ylationof activatorregardlessof nitrogenstatus.To provideacontrolsystemlackingthegalactosidepermeasepositivefeedback
loop, we simplyused theversionof the systemwiththelacY null
mutation.To providea controlsystemlackingthe positivefeedback loop of the activatormodule,but containingthe positive
feedbackloop based upon galactosidepermease,additionalgewererequired(Fig. 2B). For thispurpose,we
neticmanipulations
createda systemin whichthelacZYA promoterregion(fromthe
initiationcodon of
upstreamlacO3 sitethroughthetranslational
thelacZ gene) was fusedto theactivatorstructural
gene.Thiswas
placed intotherbsregionof the chromosome,analogousto the
and combinedincellswith
ofthegenetictoggleswitch,
positioning
theglnKp-lacZYAfusion,creatinga systemwhererepressorconwithactitrolof thelac promoterregulatedactivatorexpression,
vatorthendrivingexpressionoflacZYA.

i

Synergyof PositiveFeedbackLoops. To measure the range of
was obtained,cultures
at whichbistability
inducerconcentrations
were incubatedfor12-14 h in the absence of induceron in the
presence of saturatinginducer. The cells were then washed
PNAS | January
5,2010 | vol.107 | no. 1 | 177

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

Fig. 2. Strong bistabilitywas obtained by linkingdistinctpositive feedback loops. (A) Genetic toggle switch with a single positive feedback loop. (Left)
Schematic depiction of the genetic systemin which the genetic toggle switchdrivesthe expression of the lacZYA operon, but the lacY gene contains a null
mutation. (Right) Resultof bistabilityexperiment,showing an ~1 2-fold range of IPTGconcentrationsat which the systemdisplayed bistability.Symbols:0,
naive (uninduced) culture; Q, preinduced culture. (B) Genetic systemwith a single positive feedback loop based on galactoside permease. (Left) Schematic
of lacZYA. The lacY product,galactoside permease, provides positivefeedback byfacilitatingthe uptake of IPTG,which inactivatesrepressor.(Right)Resultof
bistabilityexperiment,showing an ~4-fold range of inducerconcentrationsat which the systemdisplayed bistability.Symbolsare as in A. (C) Genetic system
with two positive feedback loops. (Left) Schematic depiction of the genetic systemin which the genetic toggle switch drivesthe expression of the lacZYA
operon, as in A except with a wild-typelacY gene. (Right) Result of bistabilityexperimentshowing that bistabilitywas obtained over an ~480-fold range of
/
IPTG. For A-C, cells were grown in minimalmedium with succinate as the carbon source and glutamine as the nitrogensource.

and diluted 1 millionfoldinto freshmedium conthoroughly
of inducer.These cultureswere
tainingvariousconcentrations
thengrownto midlogphase and the level of the lacZ product,
ß-galactosidasewas measured(32).
The greatestcontribution
of the positivefeedbackloops is
expectedin our systemwhenthe cells are grownon mediumin
whichcataboliterepressionis minimizedand thusthegalactoside
For thispurpose,
permeaseloop is operatingwithoutinhibition.
we used succinate-basedminimalmedium.Under these conditions,the systemwitha singlepositivefeedbackloop in which
activatordrivesits own transcription
over a
displayedbistability
12-foldrangeofinducerconcentrations
with
(Fig. 24), consistent
earlierobservations(12). The controlsystemlackingthepositive
feedbackoftheactivatoron itsowntranscription,
butcontaining
a
singlepositivefeedbackloop formedby galactosidepermease,
over an ~4 -foldrange of inducerconcendisplayedbistability
trations(Fig. 2B), similarto thebistability
observedforthenative
lacZYA system(Fig. SI). Remarkably,under these same conditions,the systemwithtwo functionalpositivefeedbackloops
switch,"
(strainDE1010), whichwe referto as the"double-toggle
over an ~480-foldrangeof inducerconcendisplayedbistability
trations(Fig. 2C). Thus,thetwopositivefeedbackloops displayed
powerfulsynergyin increasingthe range of inducer concentrationsoverwhichbistability
was observed.

The bistability
of the double-toggle
switchcould be controlled
byincreasingcataboliterepression.For a verymodestdegreeof
in the succataboliterepression,
we includedcasein hydrolysate
cinate growthmedium.Under these conditions,the range of
IPTG concentrations
at whichthesystemwas bistablewas ~ 100fold (Fig. 3/4).Strongercataboliterepressionwas obtainedby
usingmediumcontainingboth glucose and casein hydrolysate;
undertheseconditionstherangeofIPTG concentrations
atwhich
the systemwas bistablewas only~25 -fold(Fig. 3B). Apparently,
cataboliterepressioncould be used to controlthecontribution
of
the galactosidepermease feedbackloop. Even thoughthe galactoside permease feedbackloop, when actingalone, did not
in thepresenceof glucose(Fig. SI),
bistability
displaysignificant
in thecontextofthedouble-toggle
switchthisloop stillincreased
therangeofbistability
about ~2-foldin mediumcontainingboth
This can be discernedby comglucose and casein hydrolysate.
parisonwiththecontrolstrainwitha nullmutationinlacY grown
underthe same conditions(Fig. 3C).
Discussion

We observed that distinctpositivefeedbackloops employing
different
biochemicalmechanismscould be linked to provide
and thatthefunctions
ofsuchsystems
powerfulgeneticbistability
were scalable byfactorsthataffectedone of thefeedbackloops.

178 | www.pnas.org/cgi/doi/10.1073/pnas.09083
14107

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

Chang et al.

Fig. 3. instabilityof the double-toggle switchwas tunable by growth substrates causing catabolite repression. (A and B) The double-toggle switchstrain
depicted in Fig.2Cwas examined in medium containingsuccinate and casein hydrolysate(>A)and in medium containingglucose and casein hydrolysate(B). (C)
The experimentis as in B, but the straincontained a null mutation in lacY and thus had only a single functioningpositive feedback loop.

could be built up piecewise,by the coherent
Thus, testability
linkage of biochemicallydistinctfeedback mechanisms.Furbecause the different
feedbackloops interactedwith
thermore,
each otherin a nonlinearway(in ourcase,bycausingan increase
in the sensitivity
of the responseto activator),fairlyweak feedback loops acted synergistically
to produceextensivebistability.
These observationprovidea generallyapplicablefoundationfor
The
therationalengineering
ofsynthetic
geneticbistablesystems.
withnumerousinteracting
prevalenceofcomplexnaturalsystems
feedbackloops, whichregulatecriticalresponsessuch as progressionthroughthecell cycle,circadianclocks,or development,
by
may reflectthis capacityto build stronggeneticbistability
collaborationof multipleweak elements.The sharpstate transitionsinvolvedin clocks,cell cycles,and irreversible
morphogenic pathwaysare likelyto requiregeneticbistabilityover a
broadrangeof conditions(e.g., refs.7 and 22), whichcould not
evolvein a singlestepand had to be builtup piecewise.
the pathTo developnovelsynthetic
geneticbistablesystems,
wayfollowedbynaturecouldbe mimickedbydevelopingcircuitry
wheredifferent
feedbackloops act coherently.
Our workshows
thatthereis no necessityforusinghighlyengineeredproteins;
affecttranscriptional
activationor represfactorsthatindirectly
sion are just as good as factorsthat directlyinteractwiththe
Thus,forexample,in ourcase
DNA, as longas theyare effective.
we wouldexpectthata circuitwhereactivatorbroughtabout the
repressionof repressorsynthesis
(alongwiththe activationof its
own synthesis)
would also bringabout verydramaticbistability.
Such a systemwouldbe closerto thatused byRossi (29) thanis
of using indirect
the systemused in this paper. The flexibility
the
methods,suchas regulationofthepermeasethatinternalizes
the
inducer,followsthe pathwayused bybacteriaand simplifies
of systems.
engineering
1. Pomerening JR, Sontag ED, Ferrell JE, Jr (2003) Building a cell cycle oscillator:
Hysteresisand bistabilityin the activation of Cdc2. Nat Cell Biol.5:346-351.
2. Sha W, et al. (2003) Hysteresisdrives cell-cycletransitions in Xenopus laevis egg
extracts.Proc Nati Acad Sci USA 100:975-980.
3. Wang X, Hao N, Dohlman HG, Eiston TC (2006) Bistability,stochasticity,and
oscillationsin the mitoqen-activatedprotein kinase cascade. BiophysJ 90:1961-1978.
4. FerrellJE,Jr(2002) Self-perpetuatingstates in signal transduction:Positivefeedback,
double-negative feedback, and bistability.CurrOpin Cell Biol 14:140-148.
5. Bagowski CP, Ferrell JE, Jr (2001) Bistabilityin the JNK cascade. Curr Biol 11:
1176-1182.
6. Dubnau D, LosickR (2006) Bistabilityin bacteria. Mol Microbiol 61:564-572.
7. TysonJJ,Chen KC, Novak B (2003) Sniffers,buzzers, toggles and blinkers:Dynamicsof
regulatoryand signaling pathways in the cell. CurrOpin Cell Biol 15:221-231.
8. Angeli D, FerrellJE,Jr,Sontag ED (2004) Detection of multistability,
bifurcations,and
hysteresisin a large class of biological positive-feedbacksystems.Proc Nati Acad Sci
USA 101:1822-1827.
9. CraciunG, Tang Y, Feinberg M (2006) Understandingbistabilityin complex enzymedrivenreaction networks.Proc Nati Acad Sci USA 103:8697-8702.

Chang et al.

We arguethatall ofthecases ofbistablesystems
withmultiple
interacting
geneticfeedbackloops noted above maybe demonthatwas
stratinga formof the same multistepultrasensitivity
studiedin signaltransduction
systems25 years ago (27, 28). A
and well-understood
limitation
in genetic
to bistability
significant
forthe critical
systemsis obtaininga sufficiently
highsensitivity
regulatorystep, such as the activator'seffectupon its own expressionin our system(8, 10-12). We expectthatthistypeof
geneticmultistepultrasensitivity
mayplay an importantrole in
geneticregulationrequiringa highkineticorderin those cases
where multiplefeedbackloops are focused on controllingthe
expressionof a gene.
Methods
Genetic Elements.The activator module of the NC12 syntheticgenetic clock
and the fusion of the glnK promoter to the lacZYA structuralgenes were
described previously(12, 30). The fusion of the lacZ promoter to the glnG
structuralgene was constructedin several steps, as described in SI Text,and
the primersused for constructionof the lacY null mutation by recombineering are listed in SI Text. All molecular cloning, PCR, P1virtransduction,
and plasmid transformationused standard techniques (33, 34). The bacterial
strainsused and their relevant genotypes are listed in Table S1.
PhysiologyExperiments.Growthmedium for bistabilityexperimentsused Wsalts (12), with added vitaminB1 (0.004% wt/vol),tryptophan(0.04% wt/vol),
and glutamine (0.2% wt/vol),and contained succinate at 0.4% wt/vol,casein
hydrolysateat 0.5% wt/vol,and glucose at 0.4% wt/vol,as indicated. IPTG
was used at 0.4 mM for overnightinductionof the cultures.ß-Galactosidase
activitywas measured using the method of Miller (32).
ACKNOWLEDGMENTS.We thank PatrickO'Brien for reviewingan earlyversion of this manuscript.This work was supported by GrantGM63642 (to AJ.
N.) from the National Institutes of Health-National Institute of General
Medical Sciences.
10. FerrellJE,Jr(2008) Feedback regulation of opposing enzymes generates robust,allor-none bistable responses. CurrBiol 18:R244-R245.
11. Xiong W, FerrellJE,Jr(2003) A positive-feedback-basedbistable 'memory module'
that governs a cell fate decision. Nature 426:460-465.
12. Atkinson MR, Savageau MA, Myers JT, Ninfa AJ (2003) Development of genetic
circuitryexhibitingtoggle switch or oscillatorybehavior in Escherichiacoli. Cell 113:

597-607.
13. Novick A, Weiner M (1957) Enzyme induction as an all-or-none phenomenon. Proc
Nati Acad Sci USA 43:553-566.
14. Ozbudak EM, Thattai M, Lim HN, Shraiman Bl, Van Oudenaarden A (2004)
Multistabilityin the lactose utilization network of Escherichia coli. Nature 427:
737-740.
1.5. Santillán M, Mackey MC (2004) Influence of catabolite repression and inducer
exclusion on the bistable behavior of the lac operon. BiophysJ 86:1282-1292.
16. Wong P, Gladney S, Keasling JD (1997) Mathematical model of the lac operon:
Inducer exclusion, catabolite repression,and diauxic growth on glucose and lactose.
Biotechnol Prog 13:132-143.

PNAS

| January 5, 2010

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

| vol. 107

| no. 1 | 179

8

17. Yildirim N, Mackey MC (2003) Feedback regulation in the lactose operon: A
mathematical modeling studyand comparison with experimental data. BiophysJ8A:
2841-2851.
18. Herzenberg LA (1959) Studies on the induction of beta-galactosidase in a cryptic
strainof Escherichiacoli. Biochim BiophysActa 31:525-538.
19. Cohn M, Horibata K (1959) Inhibitionby glucose of the induced synthesisof thebetagalactoside-enzyme systemof Escherichiacoli. Analysisof maintenance. J Bacteriol
78:601-612.
20. Inada T, Kimata K,Aiba H (1996) Mechanism responsibleforglucose-lactose diauxie in
Escherichiacoli: Challenge to the cAMP model. Genes Cells 1:293-301.
21. Hogema BM, et al. (1998) Inducerexclusion in Escherichiacoli by non-PTSsubstrates:
The role of the PEP to pyruvate ratio in determiningthe phosphorylationstate of
enzyme IIAGIc.Mol Microbio! 30:487-498.
22. Brandman O, FerrellJE,Jr,Li R, Meyer T (2005) Interlinkedfast and slow positive
feedback loops drive reliable cell decisions. Science 310:496-498.
23. Cui J,Chen C, Lu H, Sun T, Shen P (2008) Two independent positive feedbacks and
bistabilityin the Bcl-2apoptotic switch. PLoS ONE 3:e1469.
24. Thomas R, D'Ari R (1989) Biological Feedback (CRC, Boca Raton, FL).
25. Feng J,et al. (1992) Role of phosphorylatedmetabolic intermediatesin the regulation
of glutamine synthetasesynthesisin Escherichiacoli. J Bacteriol 174:6061-6070.

180

26. FerrellJE,Jr,Xiong W (2001) Bistabilityin cell signaling: How to make continuous
processes discontinuous,and reversibleprocesses irreversible.Chaos 11:227-236.
27. Goldbeter A, Koshland DE, Jr(1984) Ultrasensitivityin biochemical systemscontrolled
by covalent modification.Interplaybetween zero-order and multistepeffects.J Biol
Chem 259:14441-14447.
28. Koshland DE, Jr, Goldbeter A, Stock JB (1982) Amplificationand adaptation in
regulatoryand sensorysystems.Science 217:220-225.
29. Rossi FM, KringsteinAM, Spicher A, GuicheritOM, Blau HM (2000) Transcriptional
control: Rheostat converted to on/offswitch.Mol Cell 6:723-728.
30. AtkinsonMR, Blauwkamp TA, Bondarenko V, StuditskyV, NinfaAJ(2002) Activation
of the ginA, glnK, and nac promoters as Escherichiacoli undergoes the transition
from nitrogen excess growth to nitrogenstarvation.J Bacteriol 184:5358-5363.
31. Datsenko KA, Wanner BL (2000) One-step inactivation of chromosomal genes in
Escherichiacoli K-12 using PCR products. Proc Nati Acad Sci USA 97:6640-6645.
32. Miller JH (1972) Experimentsin Molecular Genetics (Cold Spring Harbor Lab Press,
Cold Spring Harbor, NY).
33. Sambrook J, FritschEF, Maniatis T (1989) Molecular Cloning. A Laboratory Manual
(Cold Spring Harbor Lab Press,Cold Spring Harbor, NY), 2nd Ed.
34. SilhavyTJ,Berman ML, EnquistLW (1 984) ExperimentswithGene Fusions(Cold Spring
Harbor Lab Press,Cold Spring Harbor, NY).

| www.pnas.org/cgi/doi/10.1073/pnas.0908314107

This content downloaded from 139.80.123.51 on Sat, 28 Nov 2015 12:19:27 UTC
All use subject to JSTOR Terms and Conditions

Chang et al.