Cellular Computation and Communicationsusing Engineered Genetic Regulatory Networks

Ron WeissAdvisors: Thomas F. Knight, Gerald Jay Sussman, Harold Abelson

Artificial Intelligence Laboratory, MIT

Cellular Robotics

=C

CA

B

Dgene

gene

gene

AB

CD

NAND NOT

Biochemical Logic circuit

Environment

sensors actuators

Vision

• A new substrate for engineering: living cells– interface to the chemical world– cell as a factory / robot

• Logic circuit = process description– extend/modify behavior of cells

• Challenge: engineer complex, predictable behavior

Applications

• “Real time” cellular debugger– detect conditions that satisfy logic statements

– maintain history of cellular events

• Engineered crops / farm animals– toggle switches control expression of growth hormones, pesticides

• Biomedical– combinatorial gene regulation with few inputs

– sense & recognize complex environmental conditions

• Molecular-scale fabrication– cellular robots that manufacture complex scaffolds

“Programming” Cells

plasmid = “user program”

Biochemical Inverters

signal = concentration of specific proteinscomputation = regulated protein synthesis + decay

Engineering Challenges

• Map logic circuits to biochemical reactions• Circuit design and implementation:

– conventional interfaces

– sensitivities to chemical concentrations

– understand affinities of molecules to each other

– process engineering to adjust trigger levels, gains

– CAD tools (BioSpice)

Contributions

• Experimental results:– Built and characterized a small library of logic gates

• 4 different dna-binding proteins (lacI, tetR, cI, luxR)

• 12 modifications of gates based on cI protein

• transfer functions (input/output relationship)

– Built and tested several logic circuits • combined 3 gates based on transfer functions

– Engineered communication between cells• chemical diffusions carry message

• CAD tools and program design:– BioSpice (circuit design/verification)

– Microbial Colony Language

Outline

• A Model for Programming Biological Substrates– Example: Pattern formation

– Microbial Colony language

• In-vivo digital circuits– Cellular gates: Inverter, Implies

– BioSpice circuit simulations & design

– Measuring and modifying “device physics”

• Intercellular communications

– Additional gate: AND

– BioSpice simulations & design

– Measuring “device physics”

Programming Biological Substrates

• Constraints/Characteristics:– Simple, unreliable elements– Local, unreliable communication– Elements engineered to perform tasks

• Example task: form cellular-scale patterns

Another Example: Differentiation

Cells differentiate into bands of alternating C and D type segments.

A program for creating segments:

(start Crest ((send (make-seg C 1) 3)))

((make-seg seg-type seg-index) (and Tube (not C) (not D)) ((set seg-type) (set seg-index) (send created 3)))

(((make-seg) (= 0)) Tube ((set Bottom)))

(((make-seg) (> 0)) Tube ((unset Bottom)))

(created (or C D) ((set Waiting 10)))

(* (and Bottom C 1 (Waiting (= 0))) ((send (make-seg D 1) 3)))

(* (and Bottom D 1 (Waiting (= 0))) ((send (make-seg C 2) 3)))

(* (and Bottom C 2 (Waiting (= 0))) ((send (make-seg D 2) 3)))

(* (and Bottom D 2 (Waiting (= 0))) ((send (make-seg C 3) 3)))

Microbial Colony Language (MCL)

message condition actions

How can we accomplish this?

• Boolean state variables– DNA binding proteins

• Biochemical logic circuits– genetic regulatory networks

• Intercellular signaling chemicals– enzymes that make small molecules

biocompiler: MCL genetic circuits

Outline

• Programming Biological Substrates– Pattern Formation

– Microbial Colony language

• In-vivo digital circuits– Cellular gates: Inverter, Implies

– BioSpice circuit simulations & design

– Measuring and modifying “device physics”

• Intercellular communications

– Additional gate: AND

– BioSpice simulations & design

– Measuring “device physics”

Why Digital?

• We know how to program with it– Signal restoration + modularity = robust complex circuits

• Cells do it– Phage λ cI repressor: Lysis or Lysogeny?

[Ptashne, A Genetic Switch, 1992]

– Circuit simulation of phage λ[McAdams & Shapiro, Science, 1995]

• Ultimately, combine analog &digital circuitry

Logic Circuits based on Inverters

• Proteins are the wires/signals• Promoter + decay implement the gates• NAND gate is a universal logic element:

– any (finite) digital circuit can be built!

X

Y

R1 Z

R1

R1X

Y

Z= gene

gene

gene

NAND NOT

Examples of Useful Circuits

• Logic statements:

– (x AND y AND z) OR (NOT u)

• Decoders:

– Turn ON 1 of 8 genes using only 3 inputs

• Counters

• Memory, Toggle switches

• Clocks

BioCircuit Computer-Aided Design

SPICE BioSPICE

steady state dynamics intercellular

• BioSpice: a prototype biocircuit CAD tool–simulates protein and chemical concentrations–intracellular circuits –intercellular communication

“Proof of Concept” Circuits• Work in BioSpice simulations [Weiss, Homsy, Nagpal, 1998]

• They work in vivo – Flip-flop [Gardner & Collins, 2000], Ring oscillator [Elowitz & Leibler, 2000]

• Models poorly predict their behavior

time (x100 sec)

[A]

[C]

[B]

B_S

_R

A

_[R]

[B]

_[S]

[A]

time (x100 sec)

time (x100 sec)

RS-Latch (“flip-flop”) Ring oscillator

Evaluation of the Ring Oscillator

Reliable long-term oscillation doesn’t work yet Need to match gates

[Elowitz & Leibler, 2000]

Measuring & Modifying “Device Physics”

• Why?– Different elements have widely varying characteristics

– Need to be matched

• Assembled and characterized a library of components – Constructed and measured gates using 4 genetic candidates

• lac, tet, cI, lux

– Created 12 variations of cI in order to match with lac:• modified repressor/operator affinity

• modified RBS efficiency

• other mechanisms: protein decay, promoter strength, etc..

• Established component evaluation criteria – Initially, focused on steady state behavior

Steady-State Behavior: Inverter

“ideal” transfer curve: gain (flat,steep,flat) adequate noise margins

[input]

“gain”

0 1

[output]

This curve can be achieved using proteins that cooperatively bind dna!

This curve can be achieved using proteins that cooperatively bind dna!

Measuring a Transfer Curve

• Construct a circuit that allows:– Control and observation of input protein levels– Simultaneous observation of resulting output levels

“drive” gene output gene

R YFPCFP

inverter

• Also, need to normalize CFP vs YFP

Repressors & Inducers

• Inducers that inactivate repressors:– IPTG (Isopropylthio-ß-galactoside) Lac repressor

– aTc (Anhydrotetracycline) Tet repressor

• Use as a logical Implies gate: (NOT R) OR I

operatorpromoter gene

RNAP

activerepressor

operatorpromoter gene

RNAP

inactiverepressor

inducerno transcription transcription

Repressor Inducer Output

0 0 10 1 11 0 01 1 1

RepressorInducer

Output

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Fluorescence (FL1)

Eve

nts

Drive Input Levels by Varying InducerIPTG (uM)

0

250

1000

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Eve

nts

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Eve

nts

IPTGpINV-1024125 bp

Kan(r) lacI

EYFP

P(LAC)

P(lacIq)

p15A ori

T0 Term

T1 Term

(or ECFP)

plasmid

promoter

protein coding sequence

IPTG

YFP

lacI[high]

0(Off) P(LtetO-1)

P(R)

1.00

10.00

100.00

1,000.00

0.1 1.0 10.0 100.0 1,000.0 10,000.0

IPTG (uM)

FL

1 pINV-112-R1

pINV-102

Also use for yfp/cfp calibration

Controlling Input Levels

Measuring a Transfer Curve for lacI/p(lac)

EYFPlacIP(LAC)P(LtetO-1)

RBSIIRBSII

tetRLambda P(R-O12)

RBSII

aTc

ECFP

“drive”

output

aTc

YFPlacICFP

tetR[high]0

(Off) P(LtetO-1)

P(R)

P(lac)

measure TC

Transfer Curve Data Points

01 10

1 ng/ml aTc

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Fluorescence (FL1)

Eve

nts

undefined

10 ng/ml aTc 100 ng/ml aTc

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Fluorescence (FL1)

Eve

nts

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000 10,000

Fluorescence (FL1)

Eve

nts

1

10

100

1000

1 10 100 1000

Input (Normalized CFP)

Ou

tpu

t (Y

FP)

lacI/p(lac) Transfer Curve

aTc

YFPlacICFP

tetR[high]0

(Off) P(LtetO-1)

P(R)

P(lac)

gain = 4.72gain = 4.72

Evaluating the Transfer Curve

• Noise margins:

0

200

400

600

800

1,000

1,200

1,400

1 10 100 1,000

Fluorescence

Eve

nts

30 ng/mlaTc

3 ng/mlaTc

1

10

100

1,000

0.1 1.0 10.0 100.0

aTc (ng/ml)

Flu

ore

scen

ce

• Gain / Signal restoration:

high gainhigh gain

* note: graphing vs. aTc (i.e. transfer curve of 2 gates)

10

1

102

103

100

101

102

100

101

102

103

IPTG (mM)

aTc (ng/ml)

Me

dia

n F

LR

Transfer Curve of Implies

YFPlacI

aTcIPTG

tetR[high]

Measure cI/P(R) Inverter

OR1OR2 structural gene

P(R-O12)

• cI is a highly efficient repressor

cooperativebinding

IPTG

YFPcI

CFPlacI[high]0

(Off) P(R)P(lac)

• Use lacI/p(lac) as driver

highgain

cI bound to DNA

Initial Transfer Curve for cI/P(R)

• Completely flat– Reducing IPTG no additional fluorescence

• Hard to debug!

• Process engineering: Is there a mismatch between inverters based on

lacI/p(lac) and cI/P(R)?

1.00

10.00

100.00

1,000.00

0.1 1.0 10.0 100.0 1,000.0

IPTG (uM)O

utp

ut

(YF

P)

Inverters Rely onTranscription & Translation

mRNA

ribosome

promoter

mRNAribosome

operator

translation

transcription

RNAp

Process Engineering I:Different Ribosome Binding Sites

BioSpice Simulations

RBS

translation

start

Orig: ATTAAAGAGGAGAAATTAAGCATG strongRBS-1: TCACACAGGAAACCGGTTCGATG RBS-2: TCACACAGGAAAGGCCTCGATGRBS-3: TCACACAGGACGGCCGGATG weak

1.00

10.00

100.00

1,000.00

0.1 1.0 10.0 100.0 1,000.0

IPTG (uM)

Ou

tpu

t (Y

FP

)

pINV-107/pINV-112-R1

pINV-107/pINV-112-R2

pINV-107/pINV-112-R3

Experimental Results forModified Inverter

Process Engineering II:Mutating the P(R)

BioSpice Simulations

orig: TACCTCTGGCGGTGATAmut4: TACATCTGGCGGTGATAmut5: TACATATGGCGGTGATAmut6 TACAGATGGCGGTGATA

 

OR1

Experimental Results for Mutating P(R)

1.00

10.00

100.00

1,000.00

0.1 1.0 10.0 100.0 1,000.0

IPTG (uM)

Ou

tpu

t (Y

FP

)

pINV- 107- mut4/pINV- 112- R3

pINV- 107- mut5/pINV- 112- R3

pINV- 107- mut6/pINV- 112- R3

Lessons for BioCircuit Design• Naive coupling of gates not likely to work• Need to understand “device physics”

– enables construction of complex circuits

• Use process engineering– modify gate characteristics

1.00

10.00

100.00

1,000.00

0.1 1.0 10.0 100.0 1,000.0

IPTG (uM)

Ou

tpu

t (Y

FP

)

Outline

• Programming Biological Substrates– Pattern Formation

– Microbial Colony language

• In-vivo digital circuits– Cellular gates: Inverter, Implies

– BioSpice circuit simulations & design

– Measuring and modifying “device physics”

• Intercellular communications

– Additional gate: AND

– BioSpice simulations & design

– Measuring “device physics”

Intercellular Communications

• Certain inducers useful for communications:1. A cell produces inducer

2. Inducer diffuses outside the cell

3. Inducer enters another cell

4. Inducer interacts with repressor/activator change signal

(1) (2) (3) (4)

mainmetabolism

Activators & Inducers

• Inducers can activate activators:– VAI (3-N-oxohexanoyl-L-Homoserine lacton) luxR

• Use as a logical AND gate:

operatorpromoter gene

RNAP

inactiveactivator

operatorpromoter gene

RNAP

activeactivator

inducerno transcription transcription

Output

Activator Inducer Output

0 0 00 1 01 0 01 1 1

Activator

Inducer

BioSpice: Intercellular Communications

chemicalconcentration

• Small simulation: – 4x4 grid

– 2 cells (outlined)

(1) original I = 0

(2) introduce D send msg M

(3) recv msg set I

(4) msg decays I latched

Light organ

Eupryma scolopes

Quorum Sensing

• Cell density dependent gene expression

Example: Vibrio fischeri [density dependent bioluminscence]

The lux Operon LuxI metabolism autoinducer (VAI)

luxR luxI luxC luxD luxA luxB luxE luxG

LuxR LuxI(Light)

hv(Light)

hvLuciferaseLuciferase

P

P

Regulatory Genes Structural Genes

Density Dependent Bioluminescence

free living, 10 cells/liter<0.8 photons/second/cell

symbiotic, 1010 cells/liter 800 photons/second/cell

A positive feedback circuit

luxR luxI luxC luxD luxA luxB luxE luxG

LuxRLuxI

P

P

Low Cell DensityLow Cell Density

luxR luxI luxC luxD luxA luxB luxE luxG

LuxR LuxI

(Light)hv

(Light)hvLuciferaseLuciferase

P

P

High Cell DensityHigh Cell Density

LuxRO O

O

ONH

O OO

ONH

O OO

ONH

O OO

ONH

LuxR

(+)

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

O OO

ONH

Similar Signalling Systems

N-acyl-L-Homoserine Lactone Autoinducers in Bacteria

Species Relation to Host Regulate Production of I Gene R Gene

Vibrio fischeri marine symbiont Bioluminescence luxI luxR

Vibrio harveyi marine symbiont Bioluminescence luxL,M luxN,P,Q

Pseudomonas aeruginosa Human pathogen Virulence factors lasI lasR

Rhamnolipids rhlI rhlR

Yersinia enterocolitica Human pathogen ? yenI yenR

Chromobacterium violaceum Human pathogenViolaceum production Hemolysin Exoprotease

cviI cviR

Enterobacter agglomerans Human pathogen ? eagI ?

Agrobacterium tumefaciens Plant pathogen Ti plasmid conjugation traI traR

Erwinia caratovora Plant pathogenVirulence factors Carbapenem production

expI expR

Erwinia stewartii Plant pathogen Extracellular Capsule esaI esaR

Rhizobium leguminosarum Plant symbiont Rhizome interactions rhiI rhiR

Pseudomonas aureofaciens Plant beneficial Phenazine production phzI phzR

Circuits for Controlled Sender & Receiver

pLuxI-Tet-8 pRCV-3

Fragment of pRCV-32038 bp (molecule 4149 bp)

GFP(LVA)

LuxR lux P(L)

lux P(R)

rrnB T1 rrnB T1

• Genetic networks:

• Logic circuits:

VAI VAI

Fragment of pLuxI-Tet-81052 bp (molecule 2801 bp)

LuxIP(LtetO-1) T1

aTc

luxI VAI

* E. coli strain expresses TetR (not shown)

*

VAI

LuxRGFP

tetR

aTc

00

Experimental Setup

• BIO-TEK FL600 Microplate Fluorescence Reader

• Costar Transwell microplates and cell culture inserts with permeable membrane (0.1μm pores)

• Cells separated by function:– Sender cells in the bottom well

– Receiver cells in the top well

insert

Time-Series Response to Signal

Fluorescence response of receiver (pRCV-3)

0

500

1000

1500

2000

2500

0:00 0:30 1:00 1:30 2:00

Time (hrs)

Flu

ore

sce

nce

pRCV-3 + pUC19

pRCV3 + pSND-1

pRCV-3

pRCV-3 + pRW-LPR-2

pRCV-3 + pTK-1 AI

positive control

10X VAI extra

ct

direct signalling

negative controls

Characterizing the Receiver

Response of receiver to different levels of VAI extract

0

200

400

600

800

1,000

1,200

0.1 1 10

Autoinducer Level

Max

imu

m F

luo

resc

ence

0

25,000

50,000

75,000

aTc (ng / ml)

Rec

eive

rF

luo

resc

ence

LuxTet4B9RCV Only

Controlling the Sender’s Signal Strength

Dose response of receiver cells to aTc induction of senders

receiverssenders

overlay

receivers senders

overlay

Summary

• Built, characterized, and modified a library of cellular gates (“TTL Data Book”)

• Using parts that match, built and tested several small in-vivo digital circuits

• Engineered and tested programmable intercellular communications

• BioSpice (circuit design/verification)

• Microbial Colony Language

Future Work• New programming paradigms • Bio-compiler • Additional CAD tools • Bio-fab

– Large scale circuit design, production, and testing

• Simpler & more complex organisms:– Eukaryotes

– Mycoplasmas

• Biologically inspired logic gates• Engineer multicellular organisms• Molecular scale fabrication

vs.