Genetic parts to program bacteria
Christopher A Voigt
Genetic engineering is entering a new era, where
microorganisms can be programmed using synthetic
constructs of DNA encoding logic and operational commands.
A toolbox of modular genetic parts is being developed,
comprised of cell-based environmental sensors and genetic
circuits. Systems have already been designed to be
interconnected with each other and interfaced with the control
of cellular processes. Engineering theory will provide a
predictive framework to design operational multicomponent
systems. On the basis of these developments, increasingly
complex cellular machines are being constructed to build
specialty chemicals, weave biomaterials, and to deliver
therapeutics.
Addresses
Biophysics and Chemistry & Chemical Biology, Department of
Pharmaceutical Chemistry, University of California San Francisco,
QB3 Box 2540, 1700 4th Street, San Francisco, CA 94158, USA
Corresponding author: Voigt, Christopher A (cavoigt@picasso.ucsf.edu)
Current Opinion in Biotechnology 2006, 17:548–557
This review comes from a themed issue on
Tissue and cell engineering
Edited by James L Sherley
Available online 15th September 2006
0958-1669/$ – see front matter
# 2006 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.copbio.2006.09.001
Introduction
The genome contains commands dictating how cells eat,
reproduce, communicate, move and interact with their
environment. Cells can be programmed by introducing
synthetic DNA containing new commands that instruct
the cell to perform a set of artificial tasks in series or in
parallel. These programs consist of multiple genes and
regulatory elements that function as a system composed
of sensors, circuits and converters to control biological
responses.
A rudimentary language is emerging to genetically program
bacteria [1]. Sensors have been developed that
respond to small molecules, light and temperature
[2,3
]. Genetic circuits are available that function as
inverters, logic gates, pulse generators, band pass filters
and oscillators [4,5,6
,7]. Sender and receiver components
enable cells to communicate [8
]. Based on these genetic
parts, strains of bacteria have been developed that can
communicate to form two-dimensional patterns [8
],
control their population density [9
], synthesize antimalarial
and cancer-fighting drugs [10,11], and attack malignant
cells in response to environmental cues present in a
tumor [12
] (Figure 1).
The analogy with electronic parts is useful in constructing
genetic circuits that perform signal processing tasks. However,
theanalogy is less applicable forthe design of systems
composed of many parts. Genetic parts have problems with
interference — where one part inadvertently affects
another part — because their functions are carried out
by molecular interactions and reactions that occur in the
same confined space of the cell. This imposes the restriction
that a particular genetic circuit can only be used once
in a design. Thus, the language to program cells is going to
require redundancy, or breadth, in the available parts.
A second problem is that cells are alive. They eat, grow,
avoid stress and evolve. Bacteria undergo remarkable
changes in cell state as a function of their growth stage.
The cell volume, metabolism, membrane composition,
and global regulators change in response to the growth
media and cell state. All of these factors can impact the
function of synthetic sensors and circuits. Some are more
fragile than others and recent designs have attempted to
build genetic parts whose function is as detached from the
cell state as possible. Also, evolution can effectively
‘break’ a synthetic part by introducing mutations over
many generations.
This review has been written to introduce readers to the
most robust genetic parts that have been reused in different
designs. They have been loosely divided into three
categories (Table 1 and Supplementary material). Sensors
encompass all means by which information is received by
the cell. Genetic circuits represent how information is
processed and decisions made. Actuators describe how
the circuits and sensors can be used to control processes
in the cell. The sequences and performance characteristics
for many of these parts are available at the Massachusetts
Institute of Technology (MIT) Registry of Standard Biological
Parts (http://parts.mit.edu). When given, the part
number refers to the Registry numbering system. When
available, the transfer function of a genetic part is provided.
This is an empirical measurement that describes
how the output changes as a function of the input [4,13]
(Box 1). The focus of this review is on bacteria, although
there has been much recent work in eukaryotes [14].
Sensors and inputs
Cell-based sensors can be used to identify a microenvironment,
to direct communication between cells or to
Current Opinion in Biotechnology 2006, 17:548–557 www.sciencedirect.com
Genetic parts to program bacteria Voigt 549
Figure 1
Using genetic parts to program bacteria. Programmed bacteria can (a) autonomously form spatial patterns [8
], (b) record images of light [3
],
(c) form a biofilm in response to UV light [35
], and (d) commit suicide (left-hand panel) or kill tumor cells (right-hand panel) after reaching
a critical population density [9
,12
]. Each design involves the linkage of cellular sensors to the control of biological processes, mediated
by genetic circuits. In (a), spatial patterns are formed by using a quorum sensing system to program the communication between bacteria.
The enzyme LuxI produces a small molecule (green dots) that diffuses through the cell membrane. Once the molecule accumulates to a
sufficient concentration in the media, it binds to a regulatory protein (LuxR). This regulatory protein is then connected to a pulse generator,
which controls the expression of green fluorescent protein. Thus, cells only turn green at an intermediate concentration of the signal. This forms
rings of gene expression (green, red) around the source of the signal (blue dot). In (b), bacterial photography was achieved using a light-sensing
sensor from a cyanobacterial two-component system. The protein domain that responds to light (light blue) was fused to a signal transduction
domain from E. coli (dark blue). In addition, the metabolic enzymes (green) that produce the required chromophore (pentagons) were included.
The output of the light sensor was connected to the expression of an enzyme that turns the media black. In (c), the toggle switch (yellow, magenta)
was used to control the expression of a protein that causes the bacteria to form a biofilm. One of the repressors in the toggle switch is sensitive
to UV. Thus, in the presence of UV light, the bacteria will form a biofilm. In (d), two similar quorum sensing systems are used to control
different responses as outputs. On the left, a gene is controlled that causes the cell to commit suicide (ccdB). Once the cell density reaches a
critical threshold, the cells begin to die. On the right, a gene is controlled that causes E. coli to invade malignant cells (invasin). This gene is only
turned on when there is a high concentration of bacteria. (Note that the same parts are reused in different designs and appear in Table 1 or
Supplementary material.)
www.sciencedirect.com Current Opinion in Biotechnology 2006, 17:548–557
550 Tissue and cell engineering
Table 1
DNA parts for programming bacteria.
Name Genesa
Performance Notes References
Sensors – small molecule inducers
Lac
lacI Ptrc or Ptac

Graded population induction [64]

LacI can exist in the genome or on a plasmid
Tet
tetR Ptet

Intermediate induction difficult [64]

tetR can exist in the genome or on a plasmid
Ara 
All-or-none response [65]

300-fold induction

Strain must transport (araE), but not metabolize (DaraBCD)
arabinose

Sensitive to glucose
Ara–lac 
Dual control by arabinose and IPTG [64,66]

Many mutants/detailed parameters available
Sensors – environmental inducers
Lightb
PompC cph8 ompR
pcyA ho1

Tenfold induction/high basal activity [3
]

Responds to red light

Requires E. coli RU1012 [16]

cyA/ho1 make the chromophore PCB

cph8 is chimera with EnvZ
UVc
 The wild-type cI repressor is proteolyzed in response to UV [67]
 Turns on after a UV dose of 5 J/m2
Genetic circuits – switches and logic
Inverter 
Transfer function shown with IPTG input driving
the cI repressor
[4,68
]

Can be built with other repressors (e.g. TetR and LacI)

Mutants available with different transfer functions (e.g. A, C, R)e

Can also amplify a signal
Biphasic switch 
The promoter has multiple cI binding sites with varying affinity
that either activate or repress transcription
[33
]

Pl off at both low and high input
Toggle switch
cI Ptrc Pλ lacI

All-or-none response [34]

Inputs either small molecules or promoters that
are linked to either repressor

Bistable; hysteresis in switch

Epigenetic memory
Cell–cell communicationd 
Mutants available with different transfer functions (e.g. F, C)e
[6
]
 Parallel communication using lasIR and rhlLR [68
]

LuxR can also repress promoters [69]
Genetic circuits – dynamic responses
Pulse generator
cI t0
luxR

The response can be controlled by changing the cI rbs [6
]

Is an incoherent feedforward regulatory motif [41]
Actuators
Suicide
ccdB

Kills bacterium when expressed [9
]
Biofilm
traA

Induces biofilm formation [35
]
Adhesion/invasion
invasin

Causes E. coli to adhere to and invade mammalian cells
expressing b1-integrins, including malignant cells
[12
]
au, arbitrary units; ara, arabinose; AHL, actylhomoserinelactone; aTc, anhydrotetracycline; IPTG, isopropyl-b-D-thiogalactopyranoside; PCB, phycocyanobilin;
rbs, ribosome-binding site. a
The following notation is used to describe the genetic parts. Large colored arrows and the corresponding colored circles represent
genes and their encoded proteins. The black arrows are promoters. Gray arrows represent activating interactions and lines with blunt ends are repressing
interactions. The double dotted line represents the cell membrane. Dots without black borders are small molecules. Small-molecule transporters are shown as
purple boxes in the membrane. b
The green pentagon represents the synthesis of the required chromophore PCB. The red lightning bolt shows the activation of
the light sensor with red light. c
UV light activates RecA, which leads to the proteolytic cleavage (yellow) of cI. d
LuxI (green hexagon) is an enzyme that produces
the quorum signal AHL (green dots). e
The notation (A, C, R) and (F,C) represent genetic mutants that have different transfer functions. See references for details.
Current Opinion in Biotechnology 2006, 17:548–557 www.sciencedirect.com
make the system respond to external commands. Sensors
can be wired to turn genes on or off or to directly influence
cellular behavior; for example, the direction in which it is
swimming. There are four major classes of sensors that are
commonly engineered: cytoplasmic regulatory proteins,
two-component systems, regulatory RNAs, and environment-responsive
promoters (Table 1 and Supplementary
material).
Cytoplasmic regulatory proteins
Inducible systems allow specific genes to be turned on by
adding a small molecule to the growth media. Typically,
the inducer passes through the cell membrane and binds
to a cytoplasmic regulatory protein. This either turns on
an activator or turns off a repressor, leading to the activation
or derepression of a promoter, respectively. There
are many variations of these systems that are appropriate
for different applications.
There are several key parameters that describe the transfer
function of an inducible system. First, there is the
dynamic range of the induction. This is the difference
between the basal activity in the absence of inducer and
the maximally induced state. The form of the transfer
function is also important; for example, some systems are
strongly cooperative, making it difficult to obtain intermediate
ranges of expression.
Inducible systems can also have different populationlevel
behaviors. All of the cells in a population can
behave identically and expression increases in a graded
manner as a function of inducer concentration. By contrast,
some systems produce an all-or-none response,
where a percentage of the population turn on and a
greater fraction of the cells express the gene as more
inducer is added.
Two-component systems
Two-component systems represent the most prevalent
natural sensing motif in prokaryotes [15]. Bacteria contain
many two-component systems that are simultaneously
expressed (Escherichia coli has 32) and respond to different
stimuli, such as light, temperature, touch, metals, metabolites
and chemicals. The canonical system consists of a
membrane-bound sensor which, when stimulated, phosphorylates
a response regulator. The phosphorylated regulator
then binds to promoters to activate or repress gene
expression.
The intracellular parts of two-component systems are
homologous, with similar structures and mechanisms.
This homology can be exploited to rewire the system
by genetically fusing the extracellular sensing domain to a
new, heterologous intracellular signal transduction
domain [16,17]. This was recently demonstrated through
the construction of a synthetic sensor that gives E. coli the
ability to see light [3
]. The extracellular domain of a
cyanobacterial light sensor was fused to a signal transduction
domain from E. coli, which was used to control the
expression of a gene that produces a black pigment. This
strain can record images of light projected at a twodimensional
lawn of growing bacteria (Figure 1).
A modified two-component system has been used to
sense metabolic changes [18]. Using a strain where the
cognate sensor (NRII) has been knocked out, the NRI
protein was used to sense acetyl phosphate, which
changes in response to glucose flux. This sensor was used
to optimize the production of lycopene by diverting the
carbon flux away from the toxic byproduct, acetate. The
system has also been used to maximize protein production
[19].
The bacterial chemotaxis sensing apparatus has a similar
organization to a two-component system. Chimeras have
been made between the extracellular domain of tar (a
chemosensory protein) and the intracellular domain of
EnvZ (a two-component sensor), such that chemo-attractants
can be used to regulate gene expression [17,20]. The
inverse chimeras can also be made, where the extracellular
portion of a two-component system is fused to the
intracellular portion of tar (see Supplementary material).
This enables cells to move towards the signal received by
the two-component systems; for example, a Nar–tar chimera
produced cells that move towards nitrate [21].
Mutations affecting the tar ligand-binding site have also
been shown to direct chemotaxis towards alternate amino
acids [22].
Genetic parts to program bacteria Voigt 551
Box 1 The transfer function.
The transfer function is an empirical measurement that describes
how the output changes as a function of the input [4,13]. For
example, for a light sensor, the transfer function might be the level of
gene expression as a function of light intensity [3
]. When the
transfer functions are known, they can be used to predict how
multiple connected parts will behave as a system.
There is a stochastic component to the transfer function, as was
demonstrated beautifully by Elowitz and co-workers [41
]. This
arises from cell-to-cell variation, fluctuations owing to small numbers
of molecules, and noise in transcription and translation [47,70]. When
multiple parts are connected in series, the noise from an upstream
part can influence all of the downstream processes [43
]. Stochastic
effects represent a challenge to the assembly of parts, especially
those which have multiple parts operating in series. Formal
mathematics are being developed to incorporate the stochastic
component into the transfer function [71
].
Ideally, all of the transfer functions would be measured using a
standardized strain, genetic system and reporter. This is often not
the case, so these functions are now only a qualitative guide and
cannot yet be used to quantitatively predict how multiple circuits will
function in series or the interference between circuits operating in
parallel. There is currently an effort to standardize the measurement
of part performance, which will ultimately enable computer-aided
design.
www.sciencedirect.com Current Opinion in Biotechnology 2006, 17:548–557
Maltose-binding protein (MBP) is a periplasmic protein
that interacts with tar to direct chemotaxis towards maltose.
Hellinga and co-workers have used computational
protein design to reengineer the ligand-binding pocket of
MBP to bind to unnatural chemicals, such as trinitrotoluene
(TNT) [2], L-lactate, Zn2+
[23], and a nerve agent
analog [24
]. When the engineered MBPs are coexpressed
with the tar–EnvZ chimera, gene expression
can be controlled by these chemicals.
Environment-responsive promoters
Cells change their patterns of gene expression in response
to different environmental conditions. There are several
conditions, such as pH, temperature, oxygen concentration
or UV light, for which bacteria have existing sensing
systems. Promoters that turn on under these conditions
can be used as sensory inputs. The transcription factors
acting on a promoter do not have to be known. For
example, promoters identified through microarray experiments,
where little else is known about their activation,
could be used as inputs. The transfer function of a
promoter can be characterized like an inducible system,
where gene expression is determined as a function of the
input (e.g. temperature or pH).
There are several examples where an environment-inducible
promoter has been used as an input for a genetic
circuit. Recently, Voigt and co-workers used an anaerobic
inducible promoter to create a bacterium that can invade
malignant cells in the low-oxygen microenvironment of a
tumor [12
]. Promoters involved in the bacterial heatand
cold-shock responses have also been used as temperature
sensors [25]. A common problem is that the
dynamic range of a promoter does not match the range
required to obtain an inducible phenotype. Methods to
overcome this problem are described at the end of this
review.
RNA aptamers
An aptamer is a small RNA molecule that changes conformation
when bound to a protein, peptide or small
molecule [26
]. Aptamers can be rationally fused to elements
that regulate translation, such as antisense RNAs
that inhibit translation by interfering with a ribosomebinding
site [27
]. Translation only occurs in the presence
of a ligand. Aptamers can also be used to control the
activation of other regulatory mRNA motifs, such as
ribozymes, which can regulate genes by cutting a target
transcript [28
].
RNA regulators are easily engineered as compared with
their protein counterparts and can frequently transcend
organismal boundaries [29]. In addition, they can potentially
regulate any gene through the specification of basepairing,
and their performance characteristics can be
easily fine-tuned. This enables rational design to change
the form and threshold of the transfer function [27
].
Genetic circuits
Genetic circuits enable cells to process input signals,
make logical decisions, implement memory and to communicate
with each other (Table 1 and Supplementary
material). A variety of synthetic genetic circuits have
been constructed to mimic the information-processing
capability of electronic circuits, such as inverters, logic
gates, pulse generators and oscillators. These circuits can
be used to convert a sensor output into a biological
response and used to program the cell to perform a series
of coordinated tasks.
Switches
Genetic switches are used to turn on gene expression
once an input has crossed a threshold required for activation.
A switch can also be used as an intermediary to
connect the output of a sensor to control a biological
response. A switch can be constructed using transcriptional
activators or repressors or using post-transcriptional
mechanisms, such as DNA-modifying enzymes [30
] or
riboregulators [31
] (Supplementary material). Switches
are characterized by similar performance parameters as
inducible systems: the activation threshold, the cooperativity
of the transition, and the cell-to-cell variation.
An inverter is a switch that produces a reciprocal response.
When the input to the inverter is on, then the output is
off, and vice versa. Weiss and co-workers [4] built an
inverter by linking an input promoter to the expression of
a repressor, which then turned off a downstream promoter.
The transfer function of this inverter can be varied by
using directed evolution to modify the genetic control
elements or by using different repressors [4] (http://parts.
mit.edu; e.g. BBa_Q01121).
Biphasic switches combine positive and negative regulation
so that they are only turned on by a small band of
input. A promoter can be made biphasic by introducing a
binding site where a regulator can behave as an activator
and one where it behaves like a repressor. When the
regulator has a higher affinity for the first site, then small
concentrations induce transcription and larger concentrations
repress it. The CI promoter from phage l naturally
contains this type of regulation and has been used in
synthetic applications [32,33
].
A toggle switch has been constructed using two repressors
that cross-regulate each other’s promoter [34]. The system
can exist in two states, where one or the other
repressor is fully expressed. The switch can be flipped
between states by changing the activity of a repressor,
either by directly modifying the protein or by altering its
expression. For example, the toggle switch has been
linked to quorum sensing and a UV-sensitive repressor
has been used as an input [35]. A toggle switch can also act
like a memory device. Once the switch has been latched
into one state, a large perturbation is required to switch it
552 Tissue and cell engineering
Current Opinion in Biotechnology 2006, 17:548–557 www.sciencedirect.com
into the other state. This introduces irreversibility into
the switch.
Riboswitches regulate gene expression by blocking translation
[36]. The addition of a hairpin to the 50
-end of an
mRNA transcript that overlaps the ribosome-binding site
will efficiently prevent ribosome binding [31
]. This
hairpin can be disrupted by the expression of a small
regulatory RNA. This exposes the ribosome-binding site
and activates expression. The transfer function of this
circuit can be tuned by making nucleotide substitutions
to vary the binding free energy competition between the
internal hairpin and the small RNA.
Logic gates
Logic gates are the building blocks of digital circuits.
They apply a computational operation to convert one or
more inputs into a single output. For example, the output
of an AND gate is only on when both inputs are on. By
contrast, an OR gate is on if either (or both) inputs are
on. Many forms of logic have been demonstrated in
biological circuits, including NOT, AND, OR and
NOR (NOT OR) gates [7,37]. For truly digital logic,
the inputs and outputs are either on or off (1 or 0).
Biological logic is often fuzzy, where intermediate levels
of induction are possible [38].
Two-input logic gates have been built based on the
interactions between synthetic rRNA and mRNA. Rackham
and Chin [39
] used selection to obtain orthogonal
rRNA–mRNA pairs that would only result in protein
function when both were expressed. Orthogonal pairs
do not cross-react with their endogenous counterparts,
which makes them ideal substrates for logic. This was
demonstrated by using different topologies of orthogonal
pairs to construct OR and AND gates [40].
Dynamic circuits
The circuit functions described in the previous two sections
are defined by their steady-steady state input-output
response. Circuits can also generate a dynamic
response; for example, by functioning as a pulse generator
or oscillator. A challenge in the design of dynamic circuits
is to make them robust to environmental conditions and
to reduce cell-to-cell variation.
Cascades are a common motif in natural regulatory networks.
A transcriptional cascade is formed when a chain of
transcription factors concurrently regulates each other.
Cascades can be used to temporally order the expression
of proteins, which is important in metabolic pathways
and processes involving self-assembly. Synthetic twoand
three-stage cascades have been constructed and
analyzed extensively [41
,42
]. In addition to a temporal
ordering, each stage of a cascade modifies the transfer
function and stochastic characteristics of the overall circuit
[43
].
A pulse generator can be formed by an incoherent feedforward
motif [6
,44]. A feedforward motif occurs when
the input signal is split by turning on an intermediate
transcription factor that, together with the input, regulates
a downstream process [45]. An incoherent motif
occurs when the intermediate regulator and the input
have opposite downstream effects [44]; for example,
when the input activates a repressor and together they
influence the output promoter. This can form a pulse
generator when the repressor is turned on slowly and
strongly affects the downstream promoter.
A coherent feedforward motif can function as a time delay
[46]. This motif occurs when both the input and intermediate
regulator have the same effect on a downstream
promoter. This delay can act as a filter, where short pulses
of inputs do not lead to the activation of the circuit.
Several synthetic oscillators have been constructed by
combining a ring of three repressors, a repressor and an
activator, and incorporating parts that sense metabolic
flux [47,48,49
]. The period of these oscillators varies
from 40 min to 10 h. All of the synthetic oscillators are
uncoupled and a population rapidly desynchronizes after
a few oscillations.
Cell–cell communication
Bacteria use chemical signals to communicate with each
other. In their natural setting, these quorum systems are
used to detect the cell density and to distinguish between
species [50]. Genetic circuits participating in quorum
sensing have sender and receiver components. The signal
is sent by an enzymatically synthesized small molecule
that can freely diffuse through the cell membrane. When
the quorum molecule accumulates beyond a threshold, it
binds to and activates a regulatory protein (either cytoplasmic
or a two-component system). Quorum sensing has
been used to program cell–cell communication between
bacteria [1,6
,8
,9
,12
].
Different species of bacteria communicate using variations
of the quorum molecule. If these systems function
independently (i.e. do not cross-react), then they provide
multiple channels by which bacterial communication can
be programmed [51
,52
]. Interference occurs when the
quorum molecule from different species binds non-specifically
to the regulator protein. Arnold and co-workers
have used directed evolution to change the specificity of
the regulator towards different quorum molecules
[51
,52
]. To avoid the interference problem entirely,
synthetic systems have been developed that use metabolic
products as the quorum signal [53].
Actuators: controlling cells
Cellular behavior can be controlled by using the output
from a synthetic genetic circuit to drive a natural or
transgenic response (Table 1 and Supplementary
Genetic parts to program bacteria Voigt 553
www.sciencedirect.com Current Opinion in Biotechnology 2006, 17:548–557
material). Implementing synthetic control over a biological
process is one of the least explored areas in synthetic
biology. Research has to be done to determine how to
decouple these systems from their natural inputs and to
ultimately control their function with synthetic circuits.
Sometimes this is as simple as knocking out a transcription
factor from the genome and then placing it under the
control of a new circuit. For example, Collins and coworkers
linked a UV-responsive toggle switch to a gene
that induces biofilm formation (Figure 1) [35
]. In
the presence of UV light, this strain will form a visible
biofilm. Similarly, Voigt and co-workers used a cell–cell
communication circuit to drive the expression of a protein
that enables E. coli to adhere to and invade cancer cells
[12
]. Control over cellular movement has also been
implemented by placing a single gene under inducible
control [54].
In other instances, linking synthetic sensing and logic to
cellular behavior is significantly more challenging. Often
the dynamic range of the circuitry does not match the
physiological response, either producing a constitutively
on or off phenotype. Regulatory feedback can also complicate
the control. Finally, processes that rely heavily on
core cellular processes, such as central metabolism, often
generate unintended consequences when perturbed.
Obtaining synthetic control over a complicated, multigene
function might require deconstruction of the natural
regulation and the use of synthetic regulation to control
the entire system. A step towards this goal was recently
demonstrated by refactoring and synthesizing a version of
T7 bacteriophage, which was engineered to contain simplified
regulation [55
].
Debugging and tuning complex systems
Operationally, it is difficult to connect parts to obtain a
functioning system. Two parts cannot be connected in
series when their timing and dynamic range do not match.
To overcome this problem, it is necessary to tune the
performance characteristics of one of the parts. Sometimes,
it is possible to rationally mutate a part by replacing
an operator or ribosome-binding site with an alternative
that is expected to fix the problem [6
]. A growing
database of parameterized genetic parts (http://parts.
mit.edu) is making this approach more accessible. A
successful alternative has been to use directed evolution,
where random mutagenesis is either applied across an
entire part [4] or used to target a specific region [12
,56
].
Ultimately, it might be possible to use computer-aided
design tools to accelerate the construction of a system.
Before this can happen, it is necessary to first obtain a
sufficiently large toolbox of standardized and parameterized
parts and then to develop simplified theoretical
techniques to understand how these parts will function
together. This theory may require a combination of
detailed biochemical data to characterize the inner
workings of a part with higher level, empirical relationships
that can be used to engineer the linkage between
parts. Considering promoter regulation, statistical
mechanics can be used to link the transfer function to
the thermodynamics of transcription factor binding [57
].
Computational methods can also be used to predict which
regulatory elements are the most fruitful targets for
combinatorial mutagenesis [56
].
Conclusions
DNA synthesis and cloning technology have far outstripped
our design capacity. Chemical synthesis can
now routinely build 50 kilobases of contiguous DNA at
a reasonable cost [58]. Improved transformation techniques
enable us to put this much synthetic DNA into a cell
[59
]. Sequencing the whole genome of a microbe is
becoming faster and cheaper [60]. New microfluidics
technologies are being developed that could facilitate
the rapid determination of transfer functions at the level
of individual cells [61
]. Despite all of these advances,
we are not at the point where we can design an integrated,
working system on the 50 kilobase scale. Increasing the
complexity of the designs will require improvements in
four areas: the construction of new robust parts that can be
easily interchanged; increased understanding of how to
routinely wire parts in series; the development of new
theory and computer design methods; and standardized
data sharing between laboratories.
The focus of this review has been on the individual
genetic parts that can be combined to create more complex
systems. These parts can be used to encode when,
where, how much, and under what conditions genes are
expressed. They can be used to drive the expression of
transgenic genes, to manipulate metabolism or to drive
natural processes in the cell. Cells can be programmed to
perform a series of tasks. For example, genetic circuits can
be used to control the expression of a series of metabolic
proteins at the precise time and amount required for the
maximal synthesis of a drug or energetic compound, while
diverting the flux from competing pathways [10,11,18].
Bacteria could weave complex materials by temporally
and spatially controlling the expression of biopolymer and
modifying proteins. Cells could be used as therapeutics,
programmed to find and fix medical problems in the body
[12
,62
,63]. Entire multicomponent machines and
organelles (e.g. photosystems, secretions systems, pili
and metabolic pathways) could be transferred between
species and placed under complete synthetic control.
These projects represent a revolution in genetic engineering,
where cells are programmed to undertake large
and complex tasks.
Acknowledgements
The author thanks Elizabeth Clarke, Jeff Tabor, J Christopher Anderson,
and Drew Endy for critical comments. CAV is supported by NIH PN2EY016546,
NSF BES-0547647, the Sloan Foundation, the Pew Scholars
Program, and the Sandler Family.
554 Tissue and cell engineering
Current Opinion in Biotechnology 2006, 17:548–557 www.sciencedirect.com
Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.copbio.
2006.09.001.
References and recommended reading
Papers of particular interest, published within the annual period of
review, have been highlighted as:
 of special interest
 of outstanding interest
1. Endy D: Foundations for engineering biology. Nature 2005,
438:449-453.
2. Looger LL, Dwyer MA, Smith JJ, Hellinga HW: Computational
design of receptor and sensor proteins with novel functions.
Nature 2003, 423:185-190.
3.

Levskaya A, Chevalier AA, Tabor JJ, Simpson ZB, Lavery LA,
Levy M, Davidson EA, Souras A, Ellington AD, Marcotte EM,
Voigt CA: Engineering E. coli to see light. Nature 2005,
438:441-442.
The authors recombine a cyanobacterial phytochrome with an E. coli
signal transduction system to create a strain of bacteria that can record a
two-dimensional image of projected light.
4. Yokobayashi Y, Weiss R, Arnold FH: Directed evolution of a
genetic circuit. Proc Natl Acad Sci USA 2002, 99:16587-16591.
5. Elowitz MB, Leibler S: A synthetic oscillatory network of
transcriptional regulators. Nature 2000, 403:335-338.
6.

Basu S, Mehreja R, Thiberge S, Chen M-T, Weiss R:
Spatiotemporal control of gene expression with pulsegenerating
networks. Proc Natl Acad Sci USA 2004,
101:6355-6360.
A pulse generator is developed that uses a quorum signal as an input. See
also [8
].
7. Guet CC, Elowitz MB, Hsing W, Leibler S: Combinatorial
synthesis of genetic networks. Science 2002, 296:1466-1470.
8.

Basu S, Gerchman Y, Collins CH, Arnold FH, Weiss R: A synthetic
multicellular system for programmed pattern formation.
Nature 2005, 434:1130-1134.
The Vibrio fischeri luxIR circuit is cut into pieces and used to create
‘sender’ and ‘receiver’ cells that communicate to create beautiful twodimensional
patterns.
9.

You L, Cox RS, Weiss R, Arnold FH: Programmed population
control by cell-cell communication and regulated killing.
Nature 2004, 428:868-871.
The luxIR quorum circuit is connected to the ccdB cell suicide output. This
enabled the programming of cell death after the population surpasses a
critical density.
10. Martin VJ, Pitera DJ, Withers ST, Newman JD, Keasling JD:
Engineering a mevalonate pathway in Escherichia coli for
production of terpenoids. Nat Biotechnol 2003, 27:796-802.
11. Pfeifer BA, Admiraal SJ, Gramajo H, Cane DE, Khosla C:
Biosynthesis of complex polyketides in a metabolically
engineered strain of E. coli. Science 2001, 291:1790-1792.
12.

Anderson JC, Clarke EJ, Arkin AP, Voigt CA: Environmentally
controlled invasion of cancer cells by engineered bacteria.
J Mol Biol 2006, 355:619-627.
Sensors that respond to the environmental cues in a tumor are linked to
the expression of a protein that directs the bacteria to adhere to and
invade malignant cells.
13. Weiss R, Homsy GE, Knight TR: Towards in vivo digital circuits.
DIMACS Workshop on Evolution as Computation 1999; 1: 1–18.
14. Chin JW: Modular approaches to expanding the functions of
living matter. Nat Chem Biol 2006, 2:304-311.
15. Hoch JA, Silhavy TJ: Two-Component Signal Transduction.
Washington: ASM Press; 1995.
16. Utsumi R, Brissette RE, Rampersaud A, Forst SA, Oosawa K,
Inouye M: Activation of bacterial porin gene expression by a
chimeric signal transducer in response to aspartate.
Science 1989, 245:1246-1249.
17. Kwon O, Georgellis D, Lin ECC: Rotational on-off switching of a
hybrid membrane sensor kinase Tar-ArcB in Escherichia coli.
J Biol Chem 2003, 278:13192-13195.
18. Farmer WR, Liao JC: Improving lycopene production in
Escherichia coli by engineering metabolic control.
Nat Biotechnol 2000, 18:533-537.
19. Farmer WR, Liao JC: Acetate-inducible protein overexpression
from glnAp2 promoter of Escherichia coli. Biotechnol Bioeng
2001, 75:504-509.
20. Michalodimitrakis KM, Sourjik V, Serrano L: Plasticity in amino
acid sensing of the chimeric receptor Taz. Mol Microbiol 2005,
58:257-266.
21. Ward SM, Delgado A, Gunsalus RP, Manson MD: A NarX-Tar
chimera mediates repellent chemotaxis to nitrate and nitrite.
Mol Microbiol 2002, 44:709-719.
22. Derr P, Boder E, Goulian M: Changing the specificity of a
bacterial chemoreceptor. J Mol Biol 2006, 355:923-932.
23. Dwyer MA, Looger LL, Hellinga HW: Computational design of a
Zn2+
receptor that controls bacterial gene expression.
Proc Natl Acad Sci USA 2003, 100:11255-11260.
24.

Allert M, Rizk SS, Looger LL, Hellinga HW: Computational
design of receptors for an organophosphate surrogate
of the nerve agent soman. Proc Natl Acad Sci USA 2004,
101:7907-7912.
The ligand pocket of maltose-binding protein is redesigned to bind to a
hydrolytic product of the nerve agent sarin.
25. Mujacic M, Cooper KW, Baneyx F: Cold-inducible cloning
vectors for low-temperature protein expression in Escherichia
coli: application to the production of a toxic and proteolytically
sensitive fusion protein. Gene 1999, 238:325-332.
26.

Lee JF, Hesselberth JR, Meyers LA, Ellington AD: Aptamer
database. Nucleic Acids Res 2004, 32:D95-D100.
Database of all engineered aptamers: http://aptamer.icmb.utexas.edu/.
27.

Bayer TS, Smolke CD: Programmable ligand-controlled
riboregulators of eukaryotic gene expression. Nat Biotechnol
2005, 23:337-343.
Riboregulators are constructed that turn expression on or off in the
presence of a small molecule (theophylline). The system is demonstrated
in yeast, but should also work in prokaryotes.
28.

Winkler WC, Nahvi A, Roth A, Collins JA, Breaker RR: Control of
gene expression by a natural metabolite-responsive ribozyme.
Nature 2004, 428:281-286.
A ribozyme is reported that is activated by the metabolite glucosamine-6phosphate.
Glucosamine-6-phosphate is produced by the GlmS enzyme,
the mRNA of which is cleaved by the ribozyme. This arrangement forms a
post-transcriptional negative feedback loop.
29. Werstuck G, Green MR: Controlling gene expression in living
cells through small molecule-RNA interactions. Science 1998,
282:296-298.
30.

Ham TS, Lee SK, Keasling JD, Arkin AP: A tightly regulated
inducible expression system utilizing the fim inversion
recombination switch. Biotechnol Bioeng 2006, 94:1-4.
An input induces an invertase, which flips DNA to change the orientation
of the promoter. This produces an irreversible switch with minimal
leakiness.
31.

Isaacs FJ, Dwyer DJ, Ding C, Pervouchine DD, Cantor CR,
Collins JR: Engineered riboregulators enable posttranscriptional
control of gene expression. Nat Biotechnol
2004, 22:841-847.
Hairpins are designed to sequester ribosome-binding site and, thus, to
inhibit translation. This inhibition is overcome by the transcription of a
small regulatory RNA from a second promoter.
32. Isaacs FJ, Hasty J, Cantor CR, Collins JR: Prediction and
measurement of an autoregulatory genetic module.
Proc Natl Acad Sci USA 2003, 100:7714-7719.
33.

Michalowski CB, Short MD, Little JW: Sequence tolerance of the
phage l PRM promoter: implications for evolution of gene
regulatory circuitry. J Bacteriol 2004, 186:7899-7999.
Genetic parts to program bacteria Voigt 555
www.sciencedirect.com Current Opinion in Biotechnology 2006, 17:548–557
The authors mutate the circuit controlling the lysis–lysogeny cycle of
bacteriophage l. Variants of the biphasic PRM promoter are found that
have altered transfer functions.
34. Gardner TS, Cantor CR, Collins JR: Construction of a genetic
toggle switch in Escherichia coli. Nature 2000, 403:339-342.
35.

Kobayashi H, Kaern M, Araki M, Chung K, Gardner TS, Cantor CR,
Collins JJ: Programmable cells: interfacing natural and
engineered gene networks. Proc Natl Acad Sci USA 2004,
101:8414-8419.
An excellent demonstration of the modularity and extensibility of the
genetic toggle switch. The authors link it to several different inputs
(quorum sensing, SOS response) and use it to control the formation of
a biofilm.
36. Mandal M, Breaker RR: Gene regulation by riboswitches.
Nat Rev Mol Cell Biol 2004, 5:451-463.
37. Setty Y, Mayo AE, Surette MG, Alon U: Detailed map of
cis-regulatory input function. Proc Natl Acad Sci USA 2003,
100:7702-7707.
38. Arkin A, Ross J: Computational functions in biochemical
reaction networks. Biophys J 1994, 67:560-578.
39.

Rackham O, Chin JW: A network of orthogonal ribosome*rna
pairs. Nat Chem Biol 2005, 1:159-166.
Selections are performed to create synthetic mRNA and rRNA interaction
pairs in E. coli. The pairs do not cross-react and can be used simultaneously
as control circuits.
40. Rackham O, Chin JW: Cellular logic with orthogonal ribosomes.
JACS 2005, 127:17584-17585.
41.

Rosenfeld N, Young JW, Alon U, Swain PS, Elowitz MB: Gene
regulation at the single-cell level. Science 2005, 307:1962-1965.
The biochemical parameters underlying a two-stage repressor cascade
are inferred in vivo. Transfer functions describing this network are
obtained in individual cells.
42.

Hooshangi S, Thiberge S, Weiss R: Ultrasensitivity and
noise propagation in a synthetic transcriptional cascade.
Proc Natl Acad Sci USA 2005, 10:3581-3586.
A three-stage transcriptional cascade is constructed. The transfer function
and cell-to-cell variation is measured at each stage.
43.

Pedraza JM, van Oudenaarden A: Noise propagation in gene
networks. Science 2005, 307:1965-1969.
The authors study how a noisy upstream part of a gene network will
influence the downstream parts to which it is connected.
44. Mangan S, Alon U: Structure and function of the feedforward
loop network motif. Proc Natl Acad Sci USA 2003,
100:11980-11985.
45. Shen-Orr SS, Milo R, Mangan S, Alon U: Network motifs in the
transcriptional regulation network of Escherichia coli.
Nat Genet 2002, 31:64-68.
46. Mangan S, Zaslaver A, Alon U: The coherent feedforward loop
serves as a sign-sensitive delay element in transcription
network. J Mol Biol 2003, 334:197-204.
47. Elowitz MB, Levine AJ, Siggia ED, Swain PS: Stochastic gene
expression in a single cell. Science 2002, 297:1183-1186.
48. Atkinson MR, Savigeau MA, Myers JT, Ninfa AJ: Development of
a genetic circuitry exhibiting toggle switch or oscillatory
behavior in Escherichia coli. Cell 2003, 113:597-607.
49.

Fung E, Wong WW, Suen JK, Bulter T, Lee SG, Liao JC: A
synthetic gene-metabolic oscillator. Nature 2005, 435:118-122.
An oscillating genetic circuit is constructed from transcriptional and
metabolic components. Some of the parts are from the glucose sensor
(see Supplementary material).
50. Camilli A, and Bassler, BL: Bacterial small-molecule signaling
pathways. Science 311: 1113–1116.
51.

Collins CH, Arnold FH, Leadbetter JR: Directed evolution of
Vibrio fischeri LuxR for increased sensitivity to a broad
spectrum of acyl-homoserine lactones. Mol Microbiol 2005,
55:712-723.
The authors use random mutagenesis and DNA shuffling techniques
to identify a LuxR variant that has increased specificity towards a
non-natural acyl-homoserine lactone.
52.

Collins CH, Leadbetter JR, Arnold FH: Dual selection
enhances the signaling specificity of a variant of the
quorum-sensing transcriptional activator LuxR. Nat Biotechnol
2006, 24:708-712.
The researchers use directed evolution to alter the specificity of a quorum
sensor to respond to different small molecules. This opens up multiple
channels by which cells can be programmed to communicate.
53. Bulter T, Lee S-G, Wong WW, Fung E, Connor MR, Liao JC:
Design of artificial cell-cell communication using gene and
metabolic networks. Proc Natl Acad Sci USA 2004,
101:2299-2304.
54. Blair DF, Berg HC: Restoration of torque in defective flagellar
motors. Science 1988, 242:1678-1681.
55.

Chan LY, Kosuri S, Endy D: Refactoring bacteriophage T7.
Mol Systems Biol 2005: doi: 10.1038/msb4100025.
The authors rationally reduce the regulatory complexity of the T7 genome
and build it using DNA synthesis. The refactored genome yields a viable
T7 phage.
56.

Feng X, Hooshangi S, Chen D, Li G, Weiss R, Rabitz H: Optimizing
genetic circuits by global sensitivity analysis. Biophys J 2004,
87:2195-2202.
An algorithm is developed to determine the parameters in a circuit
model that should be targeted for mutagenesis to achieve a desired
property.
57.

Bintu L, Buchler NE, Garcia HG, Gerland U, Hwa T, Kondev J,
Phillips R: Transcriptional regulation by the numbers: models.
Curr Opin Genet Dev 2005, 15:116-124.
The physiochemical basis for the transfer function of a promoter is
described.
58. Tian J, Gong H, Sheng N, Zhou X, Gulari E, Gao X, Church G:
Accurate multiplex gene synthesis from programmable DNA
microchips. Nature 2004, 432:1050-1054.
59.

Itaya M, Tsuge K, Koizumi M, Fujita K: Combining two genomes
in one cell: Stable cloning of the Synechocystis PCC6803
genome in the Bacillus subtilis 168 genome. Proc Natl Acad Sci
USA 2005, 102:15971-15976.
The entire 3.5-megabase genome of a cyanobacterium is cloned into the
B. subtilis genome.
60. Shendure J, Porreca GJ, Reppas NB, Lin X, McCutcheon JP,
Rosenbaum AM, Wang MD, Zhang K, Mitra RD, Church GM:
Accurate multiplex polony sequencing of an evolved bacterial
genome. Science 2005, 309:1728-1732.
61.

Balagadde FK, You L, Hansen CL, Arnold FH, Quake SR:
Long-term monitoring of bacteria undergoing programmed
population control in a microchemostat. Science 2005,
309:137-140.
A microchemostat with a 16 nL growth chamber is built and used to study
a population control genetic circuit in E. coli.
62.

Dane KY, Chan LA, Rice JJ, Daugherty PS: Isolation of cell
specific peptide ligands using fluorescent bacteria display
libraries. J Immunol Methods 2006, 309:120-129.
A outer membrane display system is developed and used to select for
peptides that bind specifically to breast cancer cells. Bacteria displaying
these peptides adhere tightly to their target.
63. Yu YA, Shabahang S, Timiryasova TM, Zhang Q, Beltz R,
Gentschev I, Goebel R, Szalay AA: Visualization of tumors and
metastases in live animals with bacteria and vaccinia virus
encoding light-emitting proteins. Nat Biotechnol 2004,
22:313-320.
64. Lutz R, Bujard H: Independent and tight regulation of
transcriptional units in Escherichia coli via the LacR/O, the
TetR/O and AraC/I1-I2 regulatory elements. Nucleic Acids Res
1997, 25:1203-1210.
65. Siegele DA, Hu JC: Gene expression from plasmids containing
the araBAD promoter at subsaturating inducer concentrations
represents mixed populations. Proc Natl Acad Sci USA 1997,
94:8168-8172.
66. Lutz R, Lozinski T, Ellinger T, Bujard H: Dissecting the functional
program of Escherichia coli promoters: the combined mode of
action of Lac repressor and AraC activator. Nucleic Acids Res
2001, 29:3873-3881.
556 Tissue and cell engineering
Current Opinion in Biotechnology 2006, 17:548–557 www.sciencedirect.com
67. Little JW, Shepley DP, Wert DW: Robustness of a gene
regulatory circuit. EMBO J 1999, 15:4299-4307.
68.

Karig D, Weiss R: Signal-amplifying genetic circuit enables in
vivo observation of weak promoter activation in the Rhl
quorum sensing system. Biotechnol Bioeng 2004, 89:709-718.
The Pseudomonas Rhl quorum sensing system produces a signal that is
too weak to be detected. The authors use an inverter to amplify the signal.
69. Egland KA, Greenberg EP: Conversion of the Vibrio fischeri
transcriptional activator, LuxR, to a repressor. J Bacteriol 2000,
182:805-811.
70. Ozbudak EM, Thattai M, Kurtser I, Grossman AD, van
Oudenaarden A: Regulation of noise in the expression of a
single gene. Nat Genet 2002, 31:13-14.
71.

Mettetal JT, Muzzey D, Pedraza JM, Ozbudak EM, van
Oudenaarden A: Predicting stochastic gene expression
dynamics in single cells. Proc Natl Acad Sci USA 2006,
103:7304-7309.
The authors use fluorescence data to estimate several simplified, but
predictive parameters describing the stochastic dynamics of the genetic
circuit controlling the lactose operon.
Genetic parts to program bacteria Voigt 557
www.sciencedirect.com Current Opinion in Biotechnology 2006, 17:548–557