You are on page 1of 11

Biotechnology DOI 10.1002/biot.201200085 Biotechnol. J.

2012, 7, 856866
Journal

Review

Contextualizing context for synthetic biology


identifying causes of failure of synthetic biological systems

Stefano Cardinale and Adam Paul Arkin


Physical Biosciences Division, LBNL, Department of Bioengineering, University of California, Berkeley, CA, USA

Despite the efforts that bioengineers have exerted in designing and constructing biological Received 10 FEB 2012
processes that function according to a predetermined set of rules, their operation remains funda- Revised 12 APR 2012
mentally circumstantial. The contextual situation in which molecules and single-celled or multi- Accepted 04 MAY 2012
cellular organisms find themselves shapes the way they interact, respond to the environment and
process external information. Since the birth of the field, synthetic biologists have had to grapple
with contextual issues, particularly when the molecular and genetic devices inexplicably fail to func-
tion as designed when tested in vivo. In this review, we set out to identify and classify the sources
of the unexpected divergences between design and actual function of synthetic systems and ana-
lyze possible methodologies aimed at controlling, if not preventing, unwanted contextual issues.

Keywords: Complexity Context Environment Gene expression Synthetic biology

1 Introduction inition, is the environment in which a system finds


itself. Organisms find themselves in environmental
Living systems are problem-solving systems. Even contexts where fluctuations of physical variables,
simple bacteria and viruses have solved the prob- such as temperature and osmolarity, and dynamical
lem of surviving in every environment in which life change in population density, diversity and interac-
can exist. They inhabit niches that are nearly com- tion occur. These fluctuations impact physiology
pletely isolated in the depths of the earth or in hy- and fitness in complex ways. In fact, every process
per-dense diverse, competitive communities in top in the cell is also subject to context since each de-
soils or around plant roots. It is extraordinary how pends on the life of the host via direct and indirect
robust these organisms can be to changes in their interactions with cellular resources and compo-
environment [1]. No one appreciates this capabili- nents. They are also affected by the levels of both
ty more than biological engineers who have yet to substrates and products, and the parameters of the
learn to design systems that are similarly flexible cellular milieu, including the local redox potential,
yet conserve designed function. osmolarity, porosity/viscosity/crowding, and tem-
One of the common goals of synthetic biology is perature.
to make the design of new function vastly more ef- We tend to assume that, for processes endoge-
ficient, safe, understandable, and predictable [2]. nous to the cell, most of these dependencies are
This field is likely to have a profound impact on evolutionarily optimized in some way to provide ro-
chemical, pharmaceutical and material manufac- bust and effective cellular function and thereby fit-
turing, environmental and agricultural engineer- ness. It has also been conjectured that there might
ing, and health [3, 4]. A much cited barrier to pre- also be optimization for plasticity/evolvability [5].
dictability in design is context. Context, in this def- Heterologous pathways have not had the advan-
tage of long periods of co-evolution with other cel-
lular substrates. They are generally subject to envi-
Correspondence: Prof. Adam Paul Arkin, E.O. Lawrence Berkeley National
ronments, such as bioreactors, that they have not
Laboratory, 1 Cyclotron Road, MS Stanley 922, Berkeley, CA 94720, USA
E-mail: aparkin@lbl.gov
experienced at length previously in the evolution-
ary history of the system. Thus, their function often
Abbreviation: TF, transcription factor; CRP, cAMP receptor protein; TB, ter- suffers from uncontrolled/unpredicted interac-
rific broth tions with the surrounding cellular context and en-

856 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


Biotechnol. J. 2012, 7, 856866 www.biotechnology-journal.com

vironment. Unlike the mode in natural systems, the many of the effects. We define three broad classes
designer likely does not want the system to evolve, of context: compositional context, host context, and
thus making plasticity undesirable. Further, the environmental context.
heterologous pathway itself is generally construct- Compositional context concerns the interface
ed from individual subsystems whose composition among biological components and the issues that
(physical and functional) is novel and therefore arise in their physical and functional integration
may generate spurious interactions unwanted or (Box 1a). Host context relates to the implicit de-
not predicted by the designer. We define all the im- pendence of a biological device on factors provided
plicit dependencies and spurious or unconsidered by the host organism for its operation. One conse-
interactions mentioned above as context effects. quence of this is a global coupling and competition
among implanted and endogenous functions
(Box 1b). Environmental context originates from
2 Subdividing context problems variables that are asserted outside the host and that
affect circuit function. Temperature, for example, is
Recently several authors have analyzed design a variable that can directly affect heterologous cir-
strategies to directly integrate synthetic circuits cuit parameters as well as host functions, which, to-
with endogenous cellular processes [6], or to create gether with nutritive factors, stressors, and even
systems that are more insensitive (i.e. robust) to other cells in the population, can affect the fitness
variation in conditions [7]. Problems of context are of the host and modulate the effect of heterologous
barriers to the reliable function of biological path- circuit function on that fitness (Box 1c).
ways, and they are often intermixed with problems Below we review the evidence for the mecha-
of robust and modular circuit design. In this review, nisms underlying these different types of context
we attempt to focus on the contextual mechanisms effect, how they affect circuit and host function, and
that are sources of variability/uncertainty that their implications for effective and reliable circuit
plague robust design. We also look at the contextu- design.
al violations of interface definitions that challenge
modular design of biological circuitry or circuit-
host integration, and point to possible solutions for 3 Compositional context
Compositional context is perhaps the easiest to ad-
dress in improving the design of synthetic systems.
Box 1: Sources of context dependencies Designers have a choice of what elements they use
in synthetic systems in their design and how they are physically instan-
tiated and interconnected in a host (biological or
a. Compositional context
otherwise). Thus, it may prove possible to make ju-
Physical integration of multiple regulators (e.g. transcrip-
tional), genetic devices or polypeptides [8, 9, 12, 18]. dicious choices in the elements we use to form the
Approximate knowledge of RNA folding and cis-/trans- basis set for controlling ubiquitous cellular func-
interaction of multiple regulators [13, 16]. tions, although the biological principles of how to
Functional composition of devices that leads to unex- create such sets of parts are far from clear. Here we
pected circuit failure [2023]. dissect examples of relevant mechanisms into
physical and functional composition problems.
b. Host context
Parasitic interactions between transplanted and host-
endogenous components [27, 28, 31, 33].
3.1 Physical composition
Reliance on cellular components can affect host physiol-
ogy [3538], can depend on circuit state [40] and be In genetic circuits, the activity/effect of regulatory
host specific [43]. and expressed sequences arrayed on the same
Synthetic devices depend on cellular processes such as DNA molecule depends on their precise ordering,
growth [46, 47], replication [53, 54] and partitioning at how they are linked across defined boundaries and
cell division [49, 50]. their structural interaction. Similarly, domains on
the same polypeptide or subunits within protein
c. Environmental context
complexes often have activities that are mediated
Temperature and pH can directly affect cellular functions
like transcription [57, 60, 61] or synthetic devices [62, 63]. by defined interfaces and are affected by immedi-
The host can mediate the effect of environmental context ate sequence/structural context. When those inter-
on the activity of synthetic components [70] or on popu- actions are desirable, they become design param-
lation dynamics [72]. eters. However, undesigned interactions can occur
because of the often-necessary spatial co-localiza-

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 857


Biotechnology Biotechnol. J. 2012, 7, 856866
Journal

Figure 1. Sources of context effects and


their connectivity. (A) Compositional con-
text. , Physical composition: unde-
signed direct interaction between synthetic
parts on the same molecule. , Functional
composition: unexpected effects from mod-
ule coupling, component titration or circuit
impedance. (B) Host context. *, Parasitic
interactions: undesirable interactions be-
tween synthetic and host endogenous com-
ponents. , Host resources: synthetic de-
vices often rely on host resources and ma-
chinery for functioning (R). Their produc-
tion and partition can affect device activity.
Alternatively, their depletion can affect host
physiology, which can then affect the func-
tion of synthetic components. Also, host-
specific variations of these endogenous
components can be a source of context
effects. , Host cell processes: synthetic
devices are coupled to cellular fundamental
processes such as replication, cell division
and growth. (C) Environmental context.
, Direct effects: environmental factors
such as temperature and pH can directly
impact part activity. , Host-mediated
coupling: the environment defines the cou-
pling circuit-host. For example, metabolic
load can be sustainable or not in relation to
the amount and quality of external nutrients.

tion of synthetic components, which can under- moters for the majority of tandem combinations,
mine their function. We define these unwanted, im- the combination Ptet PBAD did not function as pre-
plicit interactions as physical context (Fig. 1A). dicted. Why this combination failed was not obvi-
The meaning of cis-regulatory transcription ous, although effects on DNA looping or occlusion
factor (TF)-binding sequences in or near a pro- of TF-binding sites were proposed [9]. In fact, if
moter depends on the physical context through more complex transcriptional regulation becomes
their spatial arrangement. In one study, Cox et al. necessary, occlusion of regulator binding will prob-
[8] shuffled promoter elements to obtain a library ably gain importance as a design issue, and con-
of promoter architectures. The library generated a struction and screening of large part libraries will
large range of expression strengths and logic gat- remain popular as an approach to design. Finally,
ing from simple re-positioning of the operator se- multigene expression can still present a challenge
quences of a two-input promoter. In promoters that because of poor termination at some Class I tran-
integrated binding of two repressors, the predomi- scription terminators [10] or efficient transcription
nance of one or the other relied, often not intu- run-through in vivo [11]. These issues remain ob-
itively, on their reciprocal position relative to the ject of an intense engineering effort particularly to
transcription start site [8]. If sufficiently elucidat- improve the efficiency of terminators for T7 RNA
ed, cis-regulation at the promoter will become a polymerase [12].
powerful design parameter for more complex reg- Integrating multiple regulatory inputs on RNA
ulation. However, as noted above, currently many molecules can also be subject to undesigned inter-
spatial arrangements lead to surprising or unex- action that could undermine functionality. Recent-
pected results, making this a physical context prob- ly, Isaacs and co-workers exploited the versatility of
lem. Promoters themselves can be composed to- RNA folding and the sequence specificity of RNA
gether and tandem promoters have been used to regulation to construct pairs of riboregulators that
produce gates with an OR logic. With just two in- both repressed and activated translation from spe-
puts and one output, these OR gates could be as- cific mRNA targets in vivo [13]. In this study, the
sembled to construct all possible logic functions observed parasitic interaction involved the 5 un-
through compartmentalization of each circuit in a structured region of the engineered RNA riboregu-
different strain of Escherichia coli [9]. In this case, lator, which potentially formed alternative interac-
while little interference was present between pro- tions that interfered with the riboregulator func-

858 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


Biotechnol. J. 2012, 7, 856866 www.biotechnology-journal.com

tion [13]. In RNA-based devices, insulation of dif- output of an upstream process is a molecule (or of-
ferent functional elements can be the key for ten more formally the production rate of this mol-
restoring part behavior. For example, modification ecule) that is consumed as input by multiple down-
of structural (stem) and spacer elements [14, 15], or stream processes. To understand the effect for bio-
self-cleaving ribozymes [16], can all be used to in- chemical systems, a simple example would be a sig-
sulate RNA synthetic parts. In one study, Win and naling molecule A that is produced at a constant
co-workers introduced a cis-acting aptazyme in the rate and decays via a first order decay. Imagine two
3 untranslated region (UTR) region of the gene downstream processes that consume A to make B
target of post-transcriptional regulation [15]. In and C. The amount of A at steady state in this sys-
this case, a designed sequence spacer was used to tem is inversely proportional to the rates of con-
prevent potential interaction between the synthet- version to these molecules. Thus, the steady-state
ic RNA device and other sequences on the target rate of, say, B production depends on both these
mRNA. However, while attempting to compose at- conversion rates. If conversion to C suddenly be-
tenuators of different specificity, Lucks and co- came faster, there would be less A at steady state
workers reported that spacer sequences ranging and thus less B produced per unit time. Hence, the
from 100 to 1000 bp failed to provide sufficient in- interconnection of A production to B production is
sulation to tandem RNA attenuators, possibly be- affected by C production the latter is in the func-
cause of strong structural interactions between tional context of B and vice versa [20]. This is not
RNA elements (upstream sense RNAs and down- a desirable coupling in a modular design. Were all
stream antisense RNAs). Instead, hammerhead ri- of these systems linear, there would be a relation-
bozyme constructs were used to cleave off single ship to the concept of low-input impedance that
attenuator elements and provided the necessary could lead to fan-out failure [21]. Recently, a formal
insulation to restore functionality [16]. mathematical definition of this type of context cou-
Protein-protein interactions can be engineered pling in biochemical systems (called retroactivity),
to precisely direct signal transduction pathways by similar to the concept of impedance, has been de-
recruiting or sequestering regulators [17] or to ar- veloped [22, 23] and advances some approaches to
tificially co-localize enzymes and substrates, e.g. to mitigating these effects.
improve the yield of a biosynthesized product [18].
However, in this case, the co-localization of compo-
nents (e.g. protein domains) can also lead to un- 4 Host context
wanted context effects. When heterologous pro-
tein-protein interaction domains were added to en- The task of synthetic biologists is constrained by
zymes of the mevalonate pathway, the product titer the host organism that will contain their engi-
increased stoichiometrically with the number of neered designs. They depend explicitly or implicit-
scaffold-recruiting, peptide-ligand domains. How- ly on the host resources and machinery for func-
ever, this effect reverted when the number of scaf- tion, and ultimately their intermediate or final
folded domains increased above a certain level, products are released in a cellular context with
possibly as consequence of enzyme misfolding or which they interact. It is the implicit interactions
uncharacterized allosteric interaction caused by those aspects of host function that are assumed to
excess of ligand-peptide [18]. A precise definition be constant and unlimited for the function of a de-
of the number and relative stoichiometry of en- sign and to be insensitive to the operation of the
zymes in designed scaffolds seems to be the key to circuit that we define as host context (Fig. 1B). We
avoid such compositional issues. However, the lim- classify undesigned host interactions into two
its set by microcompartment formation, which is classes: the first includes parasitic interaction in
their most likely mechanism of function [19], might which unwanted molecular interactions between
lead to further compositional problems. components of the heterologous circuit and host
interfere with the function of both or either. The
3.2 Functional composition second class comprises functional interaction or
coupling issues. The unaccounted dependence of
Unlike physical composition, context effects in circuit function on variable and limited host re-
functional composition arise only when linking the sources as well as cellular processes induces vari-
output of one process to the input of another, and able/unpredictable circuit behavior, coupling be-
are not a direct consequence simply of physical in- tween components, or failure of circuit/host inter-
teraction. Nearly all of these result from some com- face. Of course, the coupling is a closed loop. The
petition for resources occurring between the out- functioning of the heterologous circuit can place a
put and the, possibly multiple, inputs. That is, the load on the host, affecting its fitness and consum-

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 859


Biotechnology Biotechnol. J. 2012, 7, 856866
Journal

ing resources that would normally be available for terference with host proteins of the same repressor
endogenous processes. This, in turn, can create host family. In fact, in one study at least, strain-specific
resource limitations that restrict circuit function parasitic interactions with the host, e.g. host
(e.g. by preventing optimal expression of an en- genome or proteome, have been confirmed for the
zyme). LacI family of regulators. Wang and co-workers
[31] constructed several logic circuits with orthog-
4.1 Parasitic interaction onal regulators imported from Pseudomonas sy-
ringae. This device did not function properly in five
Cellular pathways and circuits are defined, more or of seven different E. coli strains when engineered
less, by a set of molecular species, the specific in- variants of E. coli promoters Plac and Pbad were used
teractions among them, and the physical transfor- as inputs, but it worked according to design when
mations permitted by the interactions (e.g. chemi- the exogenous promoter Plux was used in place of
cal reactions). However, it is also known that en- Plac. In this study, the authors uncovered an unex-
zymes can be promiscuous [24], individual TFs can pected parasitic interaction of the lacI promoter
bind nonspecifically to DNA [25], and certain pep- with components of the LacI family of transcrip-
tide-binding domains can bind many disordered tional regulators in E. coli, which undermined cir-
target peptides with differing affinities [26]. cuit functionality.
Nonspecific binding and cross talk induce com- It is thought that protein complexes are subject-
petition for the subject molecule among the specif- ed to both positive and negative selection during
ic modules that depend on them and the nonspe- their evolution [32]. Positive selection could lead to
cific consumers. TFs with new binding specificities increased binding affinity, but this occurs at the ex-
have recently been obtained through directed evo- pense of possible synergistic, nonspecific binding
lution [27] or rational design [28, 29]. In at least one to non-cognate partners. Alternatively, negative se-
case the presence of undesigned parasitic interac- lection against binding to such unintended part-
tions of the engineered regulators with the host ners is thought to be an important driving force in
was observed. Desai and co-workers [28] used an the evolution of cognate versus non-cognate pro-
analysis of co-variation of TF and DNA target se- tein interactions, and could help to insulate signal-
quences to identify residues in the global regulator ing pathways in large paralagous protein families.
cAMP receptor protein (CRP) that confer different In an elegant study, Zarrinpar and co-workers [33]
specificities, and constructed orthogonal regulator- demonstrated the effects of negative selection in
operator pairs. Despite engineering several retar- the yeast Saccharomyces cerevisiae. The authors
geted variants of CRP, they encountered unexpect- constructed chimeras of the yeast osmosensor pro-
ed cross-talk problems with the endogenous wild- tein Sho1 with its SH3 domain replaced with other
type CRP, which was more promiscuous than the yeast as well as distant metazoans SH3 domains.
engineered regulators. In their work, Desai and co- When the authors tested the amount of specificity
workers suggested that the higher promiscuity of stored in this interaction surface, they found that
wild-type CRP towards operator sequences im- only the cognate domain recognized its peptide tar-
ported into E. coli from different species of bacteria get sequence, and none of the paralogous domains
originates from the lack of regulator-operator co- could. However, SH3 domains from other organ-
evolution in this organism [28]. Removing promis- isms did bind the target peptide with varying affini-
cuous binding or enzymatic activities by rational ties, possibly because they were not selected to not
engineering, without compromising function, may interact with the endogenous effector. The implica-
be difficult given the long evolutionary selection for tion is that using these domains, heterologously, in
these attributes [30]. S. cerevisiae would lead to undesired cross talk be-
In an attempt to engineer the binding specifici- tween the implanted system and the host. Such
ty of the lactose operon repressor LacI, Zhan and barriers to heterologous expression of genes have
co-workers [29] used the extensive structural and also been studied in a broader scope. Sorek et al.
functional knowledge of LacI protein-DNA inter- [34] attempted to pass 246,045 genes from 79 pro-
actions to design new repressor/operator pairs. karyotic genomes into E. coli and determined how
These orthogonal regulators showed limited cross many failed horizontal gene transfer. As expected,
talk and could be assembled in various logic cir- generally, proteins that are involved in fundamen-
cuits. However, the authors also noted that the ne- tal cellular processes such as transcription and
cessity for LacI repressors to oligomerize for re- translation have little propensity to transfer. In-
pression, which relies on protein-protein interac- triguingly, beside several ribosomal proteins, a
tion surfaces not involved in DNA binding, could be number of other classes of protein, including sev-
a source of undesigned cross talk and possible in- eral membrane transporters and pumps, could not

860 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


Biotechnol. J. 2012, 7, 856866 www.biotechnology-journal.com

be transferred into E. coli. Possibly, these proteins on the state of the circuit and the environment
could engage in spurious interactions with other of the cell. For example, in the yeast S. cerevisiae,
cellular functions affecting cell viability [34]. Burrill and co-workers [40] engineered a positive-
feedback device to sense the activation of DNA
4.2 Host loading: Reliance on host resources damage response and trigger a persistent output
from the circuit that functioned as memory device.
Synthetic genetic circuits must function in a cellu- Nevertheless, circuits in the ON state could still
lar environment, and normally use host resources affect the growth rate of activated versus non-acti-
and enzymes such as nucleotides, tRNAs or ribo- vated cells, with the former being outgrown and di-
somes for fulfilling their process requirements. luted in the population, which could lead to partial
Drawing from these pools of central resources can loss of the long-term memory function of the de-
impact the normal homeostasis of the host cell. Ex- vice [40]. The design of this memory switch natu-
pression of a heterologous protein can compete for rally leads to different host loading in the two states
limiting free ribosomes, which decreases the ex- since the ON state expresses a protein at a high lev-
pression of other proteins [35, 36], and even re- el and the OFF state does not. However, there are
duces the number of functional ribosomes in the alternative switch designs in which load can be bet-
cell [36]. In particular, Dong and co-workers [36] ter balanced. DNA inversion-based memory
showed that overexpression of -galactosidase switches in which state is held by the configuration
such that it comprised 30% of total cell protein com- of DNA after action of a recombinase could provide
pletely halted cell growth. A more recent example a lower load on the host [41, 42]. An invertase can
shows that a particular coding sequence can inter- flip a fragment of DNA to a different orientation, a
act with the host differentially and lead to indirect mechanism that can serve directly as memory. Al-
coupling between systems [37]. Here, the expres- ternatively, if a protein must be expressed, the frag-
sion of two of three encoded proteins varied lin- ment might contain a promoter that expresses dif-
early with plasmid copy number. The third, GFP, fell ferent proteins in each orientation, which assert
off differently at high copy number apparently due similar loads on the cell and avoid subpopulations
to specific load from expression of this sequence. of cells with different growth characteristics.
The load from gfp affected the linearity of the ex- As synthetic biologists expand the number of
pression of the other sequences when placed on hosts in which they wish to transplant synthetic de-
the same plasmid. vices, being able to predict how each circuit will
Models describing in detail the coupling be- function in a new context or environment is clear-
tween protein synthesis and cell homeostasis re- ly desirable. Nevertheless, genetic circuits can
main very important. The impact of protein overex- function differently even in different strains of the
pression on growth was recently found to be large- same species of bacteria. For example, in the work
ly a consequence of re-allocation of ribosomes, of Balagadde et al. [43], their population control cir-
constrained by a parameter describing nutritional cuit experienced unexpected oscillations over sev-
capacity [38]. Interestingly, in their model Scott and eral days before reaching a lower steady-state pop-
co-workers [38] predict the presence in E. coli of a ulation density. The cause of the oscillations was
feedback mechanism that balances the amount of not clear and, of the two different E. coli strains
cellular ribosomes, possibly triggered by changes tested, one MC4100Z1 had dampened oscilla-
in the endogenous pool of tRNAs, in response to re- tions, while the other Top10 had sustained os-
duced translational activity. The constraints im- cillations for the whole length of the experiment.
posed by host resources are beginning to set Furthermore, the two strains had a very different
bounds on the size and activity of synthetic circuits phenotypic response to the killer protein CcdB:
that may be expressed in different hosts and point MC4100Z1 showed a distinct stress phenotype,
to bottlenecks for which, in the future, we might de- while Top10 did not [43]. These strain-specific dy-
sign solutions. Importantly, the host organism also namics are clear examples of unaccounted, contex-
plays a strong role in determining which configu- tual issues that arise from interfacing a circuit with
rations of codons lead to optimal translation of the the host genetic background. However, because
protein, particularly by defining the relative abun- MC4100Z1 and Top10 are so closely related, it is
dance of each charged tRNA [39]. likely that the genetic basis of this difference in cir-
It is likely that almost all heterologous circuits cuit response to the host can be mapped, and the
place some load on the cell, thereby affecting either mechanistic causes determined.
its growth or its ability to carry out certain process- Predicting what cellular process, metabolite or
es. The significance of the load on cellular resource resource will affect the function of a device is, for
and hence on fitness is, of course, both dependent now, very difficult. In a surprising example of a load

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 861


Biotechnology Biotechnol. J. 2012, 7, 856866
Journal

issue being beneficial to circuit function, Stricker et demonstrated [47]. This is also necessary to count-
al. [44] designed a genetic oscillator that behaved er issues associated with plasmid partitioning at
more robustly and functioned at a wider range of cell division [49], and methods for controlling the
inducer concentrations than predicted. They dis- dependence of circuit function on the abundance of
covered that their circuit saturated the endogenous encoding genetic material can be useful. In partic-
proteolytic machinery, inducing coupling among ular, Bleris and co-workers showed that certain
circuit components that was critical for robust os- types of circuit topology, such as the incoherent
cillations to occur. There are likely a finite number feed-forward loop (iFFL), adapt to changes in the
of such key interaction points of heterologous cir- DNA template abundance [50].
cuit with cellular resources, and it is a key chal- Plasmid maintenance can have unexpected,
lenge to design against them or, alternatively, ex- toxic consequences on the cell such as, for instance,
ploit them. It is more complicated, perhaps, to iso- increased glucose uptake and detrimental accumu-
late a heterologous circuit from dependence on lation of metabolite intermediates [51]. As one fur-
host housekeeping machinery and metabolism. ther confirmation of the multiple, unexpected con-
One possibility that we have already mentioned is sequences of unbalanced metabolite pools, a dra-
to physically confine different subsystems of a de- matic change in charged tRNAs can interfere with
sign to different cells [9]. With this method, both plasmid replication control and lead to a dramatic
the metabolic load on the host and the interference increase in plasmid copy number [52]. Clearly,
of parasitic interactions can be limited by reducing there is a positive feedback between plasmid-asso-
the number of circuit components present in each ciated activities and host metabolism that can lead
cell type. Importantly, multicellular computing also to unregulated runaway plasmid replication (un-
has the benefit of reducing intercellular variability controlled plasmid amount) and have detrimental
with a population-wide global output [9] and al- consequences on the protein synthesis apparatus
lowing division of labor and the execution of in- or the integrity of the outer membrane [53, 54]. Ac-
compatible chemistries in the same circuit. curate control of plasmid replication can go a long
way toward limiting the metabolic burden of re-
4.3 Growth-coupled context effects combinant expression and plasmid maintenance
on the host cell. Identifying and modeling feedback
Apart from host resources, synthetic devices in a between indicators of growth and components of
cell are subject to the unavoidable recurrence of at replication regulation will be important to imple-
least three cellular processes: growth, replication ment copy-number stabilization in circuit design
and partitioning. The dependence of the cellular [55, 56].
macromolecular composition on growth rate has
been a focus of intense investigation for over
30 years [45, 46]. In some bacteria, gene copy num- 5 Environmental context
ber is strongly correlated with growth; at higher
growth rates multiple rounds of replication can si- Biological circuits and their hosts are subject to
multaneously occur. Recently, a theoretical ap- fluctuations in the environment. Resources may
proach was adopted to analyze data of protein ex- change, physical conditions such as volume and
pression from different promoters in different me- temperature may vary, and, of course, there may be
dia [47]. In this work, Klumpp and co-authors [47] other cells competing for resources in the environ-
established growth-rate dependencies of the trans- ment. In addition, the cell is continuously sensing
fer function of several types of gene circuits. Addi- the environment; therefore, various problems for
tionally, they extended the notion of growth feed- the function of a genetic circuit can originate from
back [48] identifying nonlinearity in host-circuit the very diverse interplay of host and environmen-
interaction that could lead to both expression and tal variables (Fig. 1C). We define environmental
growth bistability in absence of positive feedback. context as the unintentional coupling to variations
In this case, the growth feedback enters in the form in the environment that modifies the behavior of
of the dilution of circuit intermediates and is de- biological components either directly or indirectly
pendent on whether the circuit is encoded on a through the mediation of the host organism.
plasmid or in the host genome. At the design stage,
it is important to carefully consider whether gene 5.1 Direct effects
circuitry is implemented on a multicopy plasmid or
the host chromosome and which replication system Environmental factors can directly affect physical-
is adopted because their copy number can be dif- chemical properties of biological components and,
ferently affected by growth feedback as recently therefore, their function. Temperature can directly

862 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


Biotechnol. J. 2012, 7, 856866 www.biotechnology-journal.com

modulate transcriptional activity per se, without the ability of external resources, and, in the extreme,
involvement of trans-acting factors. For example, can lead to changes in cell fitness that can have
the promoter of bacteriophage PL shows in- deep consequence for circuit function and stability.
creased transcription when the temperature is low- Undesigned interactions among different popula-
ered [57]. Temperature-dependent changes in the tions of cells, such as cross-feeding and cross-pro-
curvature of the DNA can favor the binding of acti- tection, can lead to unexpected dynamics that can
vators [58] or repressors [59, 60]. Temperature ef- ultimately alter circuit functionality.
fects need to be accounted for during design of gene One recent example of multi-level coupling
circuits, particularly when RNA parts are used. Re- with environmental regulation arose in the attempt
cently, efforts to characterize novel synthetic RNA to engineer a propionate-regulated promoter from
thermometers were hindered by the observation Salmonella enterica. This promoter (PprpB) is acti-
that transcription/translation processes are some- vated by the regulator PrpR, which, in turn, is acti-
what more efficient at higher temperature, which vated by 2-methylcitrate derived from the metabo-
can compromise proper characterization of other lism of propionate. Additionally, the promoter is
regulatory devices as well [61]. Riboswitches have regulated by the 35-cyclic adenosine monophos-
been found to function under equilibrium condi- phate (cAMP)-CRP complex (e.g. catabolite repres-
tions, in which temperature can be an important sion) and possibly by the global regulator IHF, sug-
factor [62]. In a recent attempt to engineer ri- gesting strong environmental sensitivity of propi-
boswitches that functioned in bacteria other than E. onate-dependent transcription regulation [69]. In-
coli, Topp and co-workers [63] determined that the deed, PprpB was shown to be environment sensitive
temperature of the culture was an important vari- in a manner that depended on the particular strain
able to be considered in predicting riboswitch func- and media used for expression. The promoter
tion in the bacterium Magnetospirillum mag- showed no activity in E. coli BL21 (BLR) cells and
neticum, the optimal growth of which is at 30C. in Terrific broth (TB) medium but functioned prop-
The pH of the growth medium is another im- erly when tested in Luria broth (LB) medium. In
portant design variable in engineering biological addition, when E. coli DH1 cells were used, the pro-
parts. Probably the most investigated case is that of moter was active in both LB and TB media [70]. The
the quorum-sensing signal acyl homoserine lac- novel propionate promoter was not the only one to
tone (AHL) molecule. Non-enzymatic, tempera- show strong host-mediated environmental effect.
ture- and pH-dependent hydrolysis of AHL is no- For example, the long-established lac promoter
ticeable in aqueous solutions [64] (for a review see was repressed by the addition of 1% glucose to the
[65]). When the response to pH is predictable, this medium, which is known to activate catabolite re-
parameter can be used to effectively modulate cir- pression, and showed leaky expression in TB [70].
cuit output. For example, the steady-state cell den- These data demonstrated that complex regulatory
sity reached by a bacterial population engineered mechanisms used by bacteria to deal with nutrient
with a population-control device could be tuned by availability, such as catabolite repression, could
changing the pH of the growth medium [66]. How- have strong and unexpected impacts on synthetic
ever, pH affected other parameters in the model gene circuits when probed outside the experimen-
such as cell growth and, unexpectedly, the toxicity tal conditions used for optimization.
of the killer protein CcdB [66]. Microbes also modify their local environment
by secreting metabolites and signals, originating
5.2 Host-mediated environmental effect complex subpopulation and subpopulation/inter-
species dependencies through cross-feeding and
Host resources and subsystem activity vary with cross-protection. Cross-feeding, trade-offs in
the environment. The accessibility of common growth-phase parameters, and spatial constrains
metabolite pools, the energy charge, and the avail- within the culture, which affect the dilution of
ability of polymerases and ribosomes are all great- shared agents, are important factors involved in the
ly affected by growth phase and available external emergence and maintenance of polymorphic sub-
nutrients [46, 67, 68]. Global regulators such as populations of bacteria (reviewed in [71]). Recent-
H-NS, IHF and Lrp in E. coli can change DNA ac- ly, during growth in a bioreactor, a polymorphic
cessibility in ways that impact circuit expression; subpopulation of cells resistant to antibiotics such
proteases change their levels and thereby their as norfloxacin or gentamicin was found to protect
ability to be saturated; and cells change volume, the majority of non-resistant cells by secreting the
shape and other parameters that might affect cir- signaling molecule indole into the media, a mole-
cuit function. Likewise, the load imposed on a cell cule involved in stress tolerance in E. coli [72]. The
by a heterologous circuit is affected by the avail- design of synthetic devices with phases that can

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 863


Biotechnology Biotechnol. J. 2012, 7, 856866
Journal

differently impact fitness trade-offs between cell


subpopulations, such as a host-integrated memo- Dr. Stefano Cardinale is a postdoctoral
ry/non-memory states [40] (often encoded on plas- researcher at Lawrence Berkeley Na-
mids that can lead to the emergence of polymor- tional Laboratory and University of Cali-
phic, cooperative cells or cheaters [73]), will have fornia, Berkeley (Berkeley, USA). Dur-
to rely on careful analysis of complex population ing his PhD program at the University
dynamics to engineer safer device containments of Edinburgh (Scotland, UK), Wellcome
and evolutionary and functional robustness during Trust Center for Cell Biology, he worked
prolonged culture. on the construction, manipulation and
application of Synthetic Human Chro-
mosomes in human tissue cells. This work led to two patent applica-
6 Concluding remarks tions and several scientific publications. In his carrier, Dr. Cardinale
has received an EMBO Short-Term Fellowship to pursue research on
Synthetic devices are extraneous elements that are nuclear structures in human cells and a Darwin Trust Scholarship for
implanted in a cellular system that has used mil- his graduate studies. His current research focuses on the effect of
lions of years of evolutionary tuning to reach equi- each gene of the E. coli genome on cellular physiology and the activity
librium. Currently, we cannot map genomes to ac- of synthetic genetic circuits to identify key sources of failure of synthet-
count for all parasitic interactions that may occur ic devices in bacteria and to improve predictable design of synthetic
between the vast array of synthetic biological parts systems. His research interests are Systems and Synthetic Biology,
being built and the organisms in which they will and biotechnology.
function. Neither we will probably, in the near fu-
ture, be able to precisely model production and
partitioning of all cellular resources in all different
growth conditions for many types of organisms. Prof. Adam Arkin, Director of the Berke-
However, as synthetic biologists expand the num- ley Synthetic Biology Institute and of the
ber of hosts in which they wish to transplant syn- Physical Biosciences Division at
thetic devices, especially to more complex biologi- Lawrence Berkeley National Laboratory
cal systems found in higher eukaryotes, being able (LBNL), is the Dean A. Richard Newton
to predict how a circuit will function in a new con- Memorial Professor in UC Berkeleys
text is clearly desirable. Context dependencies Department of Bioengineering. He is
need to be part of the standard characterization of also the Chief Executive and Science
biological parts and some proposals to this end Officer of the Department of Energy
have already been made [74]. However, it remains Systems Biology Knowledgebase, technical co-manager of the large
an open challenge to create the physical map of scale Ecosystems and Networks Integrated with Genes and Molecular
where and how such dependencies arise with sim- Assemblies (ENIGMA) program, Director of Bioinformatics at the Joint
ilar precision to that found in electrical or mechan- Bioenergy Institute, and Co-director of the BIOFAB (International Open
ical engineering. Results from genome-scale stud- Facility Advancing Biotechnology). His research centers on uncovering
ies on gene-phenotype interaction suggest that, al- the evolutionary design principles of cellular networks and populations,
though host and immediate context effects might and exploiting them for applications. He and colleagues are developing
be complex, they are finite [75, 76], and could like- a framework to facilitate applications in the fields of health, environ-
ly be dealt with by formal design. In fact, as the ment, and bioenergy by combining comparative functional genomics
studies cited above have demonstrated, diagnosing and genetics, quantitative measurement of cellular dynamics, systems
the failures of designed systems to meet specifica- biological modeling of cellular networks, and cellular design. He has
tion is leading to deeper understanding of different been a member of the UC Berkeley faculty since 1999, and earned his
types of molecular mechanisms, the nature and PhD in physical chemistry from Massachusetts Institute of Technology
limitations of cellular resources, and the indirect (MIT). He was named a Fellow of the American Academy of Microbiol-
coupling among cellular subsystems, and has hint- ogy in 2007, and has been profiled in Time Magazine as a future inno-
ed at undiscovered processes. vator. He was a member of the founding class of the MIT Technology
Recently, it has been pointed out that it may be Review Young Innovators program. He is the author or co-author of
possible to design genetic circuit probes to pre- over 190 scientific papers and reviews.
cisely dissect these new cellular processes [6]. Fur-
ther, new theories of robust design and design
frameworks for biological insulation elements are and characterize these contextual effects; hence,
emerging in the toolkits of designers [7]. The syn- the field is ready to systematically address these is-
thetic biology community does not lack engineered sues and improve predictability in the design of bi-
biological parts or experimental tools to identify ological systems.

864 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


Biotechnol. J. 2012, 7, 856866 www.biotechnology-journal.com

The authors would like to thank Jeffrey Skerker, [17] Bashor, C. J., Helman, N. C., Yan, S., Lim, W. A., Using engi-
Chang Liu and Vivek Mutalik for proofreading and neered scaffold interactions to reshape MAP kinase path-
way signaling dynamics. Science 2008, 319, 15391543.
commenting the manuscript. The work described in
[18] Dueber, J. E., Wu, G. C., Malmirchegini, G. R., Moon, T. S. et
this report was funded by the Laboratory Directed al., Synthetic protein scaffolds provide modular control over
Research and Development program titled: Predic- metabolic flux. Nat. Biotechnol. 2009, 27, 753759.
tive High-Throughput Assembly of Synthetic Biolog- [19] Lee, H., Deloache, W. C., Dueber, J. E., Spatial organization of
ical Systems: From Gene Expression to Carbon Se- enzymes for metabolic engineering. Metab. Eng. 2012, 14,
questration of the U.S. Department of Energy (DOE) 242251.
under Contract No. DE-AC02-05CH11231. [20] Arkin, A. P., Signal processing by biochemical reaction net-
works, in: Walleczek, J. (Ed.), Self-Organized Biological Dy-
namics and Nonlinear Control, Cambride University Press,
The authors declare no conflict of interest. Cambridge 2000.
[21] Kim, K. H., Sauro, H. M., Fan-out in gene regulatory net-
works. J. Biol. Eng. 2010, 4, 16.
7 References [22] Saez-Rodriguez, J., Gayer, S., Ginkel, M., Gilles, E. D., Auto-
matic decomposition of kinetic models of signaling net-
[1] Pikuta, E. V., Hoover, R. B., Tang, J., Microbial extremophiles works minimizing the retroactivity among modules. Bioin-
at the limits of life. Crit. Rev. Microbiol. 2007, 33, 183209. formatics 2008, 24, i213219.
[2] Lucks, J. B., Qi, L., Whitaker, W. R., Arkin, A. P., Toward scal- [23] Del Vecchio, D., Ninfa, A. J., Sontag, E. D., Modular cell biol-
able parts families for predictable design of biological cir- ogy: retroactivity and insulation. Mol. Sys. Biol. 2008, 4, 161.
cuits. Curr. Opin. Microbiol. 2008, 11, 567573. [24] de Groot, M. J., van Berlo, R. J., van Winden, W. A., Verheijen,
[3] Khalil, A. S., Collins, J. J., Synthetic biology: applications P. J. et al., Metabolite and reaction inference based on en-
come of age. Nat. Rev. Genet. 2010, 11, 367379. zyme specificities. Bioinformatics 2009, 25, 29752982.
[4] Cambray, G., Mutalik, V. K., Arkin, A. P., Toward rational de- [25] Kalodimos, C. G., Biris, N., Bonvin, A. M., Levandoski, M. M.
sign of bacterial genomes. Curr. Opin. Microbiol. 2011, 14, et al., Structure and flexibility adaptation in nonspecific and
624630. specific protein-DNA complexes. Science 2004, 305, 386
[5] Lozada-Chavez, I., Janga, S. C., Collado-Vides, J., Bacterial 389.
regulatory networks are extremely flexible in evolution. Nu- [26] Schreiber, G., Keating, A. E., Protein binding specificity ver-
cleic Acids Res. 2006, 34, 34343445. sus promiscuity. Curr. Opin. Struct. Biol. 2011, 21, 5061.
[6] Nandagopal, N., Elowitz, M. B., Synthetic biology: Integrat- [27] Krueger, M., Scholz, O., Wisshak, S., Hillen, W., Engineered
ed gene circuits. Science 2011, 333, 12441248. Tet repressors with recognition specificity for the tetO-
[7] Randall, A., Guye, P., Gupta, S., Duportet, X., Weiss, R., De- 4C5G operator variant. Gene 2007, 404, 93100.
sign and connection of robust genetic circuits. Methods En- [28] Desai, T. A., Rodionov, D. A., Gelfand, M. S., Alm, E. J., Rao, C.
zymol. 2011, 497, 159186. V., Engineering transcription factors with novel DNA-bind-
[8] Cox, R. S., 3rd, Surette, M. G., Elowitz, M. B., Programming ing specificity using comparative genomics. Nucleic Acids
gene expression with combinatorial promoters. Mol. Syst. Res. 2009, 37, 24932503.
Biol. 2007, 3, 145. [29] Zhan, J., Ding, B., Ma, R., Ma, X. et al., Develop reusable and
[9] Tamsir, A., Tabor, J. J., Voigt, C. A., Robust multicellular com- combinable designs for transcriptional logic gates. Mol. Sys.
puting using genetically encoded NOR gates and chemical Biol. 2010, 6, 388.
wires. Nature 2011, 469, 212215. [30] Carbonell, P., Lecointre, G., Faulon, J. L., Origins of specifici-
[10] Macdonald, L. E., Durbin, R. K., Dunn, J. J., McAllister, W. T., ty and promiscuity in metabolic networks. J. Biol. Chem.
Characterization of two types of termination signal for bac- 2011, 286, 4399444004.
teriophage T7 RNA polymerase. J. Mol. Biol. 1994, 238, 145 [31] Wang, B., Kitney, R. I., Joly, N., Buck, M., Engineering modu-
158. lar and orthogonal genetic logic gates for robust digital-like
[11] Du, L., Gao, R., Forster, A. C., Engineering multigene ex- synthetic biology. Nat. Commun. 2011, 2, 508.
pression in vitro and in vivo with small terminators for T7 [32] Wang, Y., Franzosa, E. A., Zhang, X. S., Xia, Y., Protein evolu-
RNA polymerase. Biotechnol. Bioeng. 2009, 104, 11891196. tion in yeast transcription factor subnetworks. Nucleic Acids
[12] Du, L., Villarreal, S., Forster, A. C., Multigene expression in Res. 2010, 38, 59595969.
vivo: Supremacy of large versus small terminators for T7 [33] Zarrinpar, A., Park, S. H., Lim, W. A., Optimization of speci-
RNA polymerase. Biotechnol. Bioeng. 2012, 109, 10431050. ficity in a cellular protein interaction network by negative
[13] Isaacs, F. J., Dwyer, D. J., Ding, C., Pervouchine, D. D. et al., En- selection. Nature 2003, 426, 676680.
gineered riboregulators enable post-transcriptional control [34] Sorek, R., Zhu, Y., Creevey, C. J., Francino, M. P. et al.,
of gene expression. Nat. Biotechnol. 2004, 22, 841847. Genome-wide experimental determination of barriers to
[14] Bayer, T. S., Smolke, C. D., Programmable ligand-controlled horizontal gene transfer. Science 2007, 318, 14491452.
riboregulators of eukaryotic gene expression. Nat. Biotech- [35] Vind, J., Sorensen, M. A., Rasmussen, M. D., Pedersen, S.,
nol. 2005, 23, 337343. Synthesis of proteins in Escherichia coli is limited by the
[15] Win, M. N., Smolke, C. D., A modular and extensible RNA- concentration of free ribosomes. Expression from reporter
based gene-regulatory platform for engineering cellular genes does not always reflect functional mRNA levels.
function. Proc. Natl. Acad. Sci. USA 2007, 104, 1428314288. J. Mol. Biol. 1993, 231, 678688.
[16] Lucks, J. B., Qi, L., Mutalik, V. K., Wang, D., Arkin, A. P., Ver- [36] Dong, H., Nilsson, L., Kurland, C. G., Gratuitous overexpres-
satile RNA-sensing transcriptional regulators for engineer- sion of genes in Escherichia coli leads to growth inhibition
ing genetic networks. Proc. Natl. Acad. Sci. USA 2011, 108, and ribosome destruction. J. Bacteriol. 1995, 177, 14971504.
86178622.

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 865


Biotechnology Biotechnol. J. 2012, 7, 856866
Journal

[37] Hajimorad, M., Gray, P. R., Keasling, J. D., A framework and [58] Perez-Martin, J., Espinosa, M., Correlation between DNA
model system to investigate linear system behavior in Es- bending and transcriptional activation at a plasmid promot-
cherichia coli. J. Biol. Eng. 2011, 5, 3. er. J. Mol. Biol. 1994, 241, 717.
[38] Scott, M., Gunderson, C. W., Mateescu, E. M., Zhang, Z., Hwa, [59] Prosseda, G., Falconi, M., Giangrossi, M., Gualerzi, C. O. et
T., Interdependence of cell growth and gene expression: al., The virF promoter in Shigella: More than just a curved
Origins and consequences. Science 2010, 330, 10991102. DNA stretch. Mol. Microbiol. 2004, 51, 523537.
[39] Elf, J., Nilsson, D., Tenson, T., Ehrenberg, M., Selective charg- [60] Madrid, C., Nieto, J. M., Paytubi, S., Falconi, M. et al., Tem-
ing of tRNA isoacceptors explains patterns of codon usage. perature- and H-NS-dependent regulation of a plasmid-
Science 2003, 300, 17181722. encoded virulence operon expressing Escherichia coli he-
[40] Burrill, D. R., Silver, P. A., Synthetic circuit identifies sub- molysin. J. Bacteriol. 2002, 184, 50585066.
populations with sustained memory of DNA damage. Genes [61] Neupert, J., Karcher, D., Bock, R., Design of simple synthet-
Dev. 2011, 25, 434439. ic RNA thermometers for temperature-controlled gene ex-
[41] Friedland, A. E., Lu, T. K., Wang, X., Shi, D. et al., Synthetic pression in Escherichia coli. Nucleic Acids Res. 2008, 36, e124.
gene networks that count. Science 2009, 324, 11991202. [62] Lynch, S. A., Desai, S. K., Sajja, H. K., Gallivan, J. P., A high-
[42] Ham, T. S., Lee, S. K., Keasling, J. D., Arkin, A. P., Design and throughput screen for synthetic riboswitches reveals mech-
construction of a double inversion recombination switch for anistic insights into their function. Chem. Biol. 2007, 14,
heritable sequential genetic memory. PloS One 2008, 3, 173184.
e2815. [63] Topp, S., Reynoso, C. M., Seeliger, J. C., Goldlust, I. S. et al.,
[43] Balagadde, F. K., You, L., Hansen, C. L., Arnold, F. H., Quake, Synthetic riboswitches that induce gene expression in di-
S. R., Long-term monitoring of bacteria undergoing pro- verse bacterial species. Appl. Environ. Microbiol. 2010, 76,
grammed population control in a microchemostat. Science 78817884.
2005, 309, 137140. [64] Yates, E. A., Philipp, B., Buckley, C., Atkinson, S. et al.,
[44] Stricker, J., Cookson, S., Bennett, M. R., Mather, W. H. et al., A N-Acylhomoserine lactones undergo lactonolysis in a pH-,
fast, robust and tunable synthetic gene oscillator. Nature temperature-, and acyl chain length-dependent manner
2008, 456, 516519. during growth of Yersinia pseudotuberculosis and Pseudo-
[45] Churchward, G., Bremer, H., Young, R., Macromolecular monas aeruginosa. Infect. Immun. 2002, 70, 56355646.
composition of bacteria. J. Theor. Biol. 1982, 94, 651670. [65] Wang, Y. J., Huang, J. J., Leadbetter, J. R., Acyl-HSL signal de-
[46] Bremer, H. a. D., P., Modulation of Chemical Composition and cay: Intrinsic to bacterial cell-cell communications. Adv.
Other Parameters of the Cell by Growth Rate, American Soci- Appl. Microbiol. 2007, 61, 2758.
ety for Microbiology, Washington DC 1996. [66] You, L., Cox, R. S., 3rd, Weiss, R., Arnold, F. H., Programmed
[47] Klumpp, S., Zhang, Z., Hwa, T., Growth rate-dependent population control by cell-cell communication and regulat-
global effects on gene expression in bacteria. Cell 2009, 139, ed killing. Nature 2004, 428, 868871.
13661375. [67] Liang, S. T., Xu, Y. C., Dennis, P., Bremer, H., mRNA compo-
[48] Tan, C., Marguet, P., You, L., Emergent bistability by a sition and control of bacterial gene expression. J. Bacteriol.
growth-modulating positive feedback circuit. Nat. Chem. 2000, 182, 30373044.
Biol. 2009, 5, 842848. [68] Zhang, X., Liang, S. T., Bremer, H., Feedback control of ribo-
[49] Guido, N. J., Wang, X., Adalsteinsson, D., McMillen, D. et al., some synthesis in Escherichia coli is dependent on eight
A bottom-up approach to gene regulation. Nature 2006, 439, critical amino acids. Biochimie 2006, 88, 11451155.
856860. [69] Lee, S. K., Newman, J. D., Keasling, J. D., Catabolite repres-
[50] Bleris, L., Xie, Z., Glass, D., Adadey, A. et al., Synthetic inco- sion of the propionate catabolic genes in Escherichia coli
herent feedforward circuits show adaptation to the amount and Salmonella enterica: Evidence for involvement of the
of their genetic template. Mol. Sys. Biol. 2011, 7, 519. cyclic AMP receptor protein. J. Bacteriol. 2005, 187, 27932800.
[51] Diaz-Ricci, J. C., Tsu, M., Bailey, J. E., Influence of expression [70] Lee, T. S., Krupa, R. A., Zhang, F., Hajimorad, M. et al.,
of the pet operon on intracellular metabolic fluxes of Es- BglBrick vectors and datasheets: A synthetic biology plat-
cherichia coli. Biotechnol. Bioeng. 1992, 39, 5965. form for gene expression. J. Biol. Eng. 2011, 5, 12.
[52] Wegrzyn, G., Wegrzyn, A., Is tRNA only a translation factor [71] Rainey, P. B., Buckling, A., Kassen, R., Travisano, M., The
or also a regulator of other processes? J. Appl. Genet. 2008, emergence and maintenance of diversity: Insights from ex-
49, 115122. perimental bacterial populations. Trends Ecol. Evol. 2000,
[53] Ryan, W., Parulekar, S. J., Recombinant protein synthesis and 15, 243247.
plasmid instability in continuous cultures of Escherichia coli [72] Lee, H. H., Molla, M. N., Cantor, C. R., Collins, J. J., Bacterial
JM103 harboring a high copy number plasmid. Biotechnol. charity work leads to population-wide resistance. Nature
Bioeng. 1991, 37, 415429. 2010, 467, 8285.
[54] Togna, A. P., Shuler, M. L., Wilson, D. B., Effects of plasmid [73] Nogueira, T., Rankin, D. J., Touchon, M., Taddei, F. et al., Hor-
copy number and runaway plasmid replication on overpro- izontal gene transfer of the secretome drives the evolution
duction and excretion of beta-lactamase from Escherichia of bacterial cooperation and Virulence. Curr. Biol. 2009, 19,
coli. Biotechnol. Prog. 1993, 9, 3139. 16831691.
[55] Klumpp, S., Growth-rate dependence reveals design princi- [74] Canton, B., Labno, A., Endy, D., Refinement and standardi-
ples of plasmid copy number control. PloS One 2011, 6, zation of synthetic biological parts and devices. Nat.
e20403. Biotechnol. 2008, 26, 787793.
[56] Kittleson, J. T., Cheung, S., Anderson, J. C., Rapid optimiza- [75] Nichols, R. J., Sen, S., Choo, Y. J., Beltrao, P. et al., Phenotypic
tion of gene dosage in E. coli using DIAL strains. J. Biol. Eng. landscape of a bacterial cell. Cell 2011, 144, 143156.
2011, 5, 10. [76] Deutschbauer, A., Price, M. N., Wetmore, K. M., Shao, W. et al.,
[57] Giladi, H., Goldenberg, D., Koby, S., Oppenheim, A. B., En- Evidence-based annotation of gene function in Shewanella
hanced activity of the bacteriophage lambda PL promoter at oneidensis MR-1 using genome-wide fitness profiling
low temperature. FEMS Microbiol. Rev. 1995, 17, 135140. across 121 conditions. PLoS Genet. 2011, 7, e1002385.

866 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim