basic science through engineering synthetic modeling and the idea of biology inspired engineering...

7/24/2019 Basic Science Through Engineering Synthetic Modeling and the Idea of Biology Inspired Engineering 2013 Studies i

1/12

Basic science through engineering? Synthetic modeling and the idea of

biology-inspired engineering

Tarja Knuuttila a, Andrea Loettgers b

a University of Helsinki, Fabianinkatu 24 (P.O. Box 4), 00014, Finlandb California Institute of Technology, 1200 E. California Blvd., MC 114-96, Pasadena, CA 91125, USA

a r t i c l e i n f o

Article history:

Available online 17 April 2013

Keywords:

Synthetic modeling

TheRepressilator

Engineering

Integration

Design principles

Network motifs

a b s t r a c t

Synthetic biology is often understood in terms of the pursuit for well-characterized biological parts to

create synthetic wholes. Accordingly, it has typically been conceived of as an engineering dominated

and application oriented field. We argue that the relationship of synthetic biology to engineering is far

more nuanced than that and involves a sophisticated epistemic dimension, as shown by the recent prac-

tice of synthetic modeling. Synthetic models are engineered genetic networks that are implanted in a nat-

ural cell environment. Their construction is typically combined with experiments on model organisms as

well as mathematical modeling and simulation. What is especially interesting about this combinational

modeling practice is that, apart from greater integration between these different epistemic activities, it

has also led to the questioning of some central assumptions and notions on which synthetic biology is

based. As a result synthetic biology is in the process of becoming more biology inspired.

2013 Elsevier Ltd. All rights reserved.

When citing this paper, please use the full journal title Studies in History and Philosophy of Biological and Biomedical Sciences

1. Lego bricks . . .over and over again?

On the cover of the September 2011 issue ofSciencethat is ded-

icated to synthetic biology, grayish agglomerations depicted as

consisting of Lego-bricks float in the lilac-black space. The picture,

belonging to a flourishing genre of Lego-inspired synthetic biology

pictures, was produced by Equinox Graphicsa company special-

ized in producing science and engineering images. On page 1193

of the special issue one finds the following description of what is

depicted on the picture: Bacteria constructed from toy bricks rep-

resent the potential of synthetic biology to design and constructgenetic modules that can be used to introduce new functions into

existing organisms or even to engineer new biological systems.

(seeFig. 1).

The Lego-brick has become almost the epitome of synthetic

biology, understood as a kind of engineering science, which pur-

sues designing and constructing well-characterized biological

parts to create synthetic wholes. It is displayed by the goal of cre-

atingstandardbiological parts, BioBricks,1 that are interchangeable

sequences of DNA with specific function, and in an attempt to

apply to biology the practices of standardization, decoupling, and

abstraction (Endy, 2005). While this understanding of synthetic

biology, built on a metaphor of construction, has received the most

popular attentionas also the cover of the recent issue ofScience

showsthe place and role of engineering in synthetic biology is

much more nuanced than that. In fact the cover ofScience actually

contradicts some of the most central points presented in the articles

of the issue. Firstly, Nandagopal and Elowitz (2011, pp. 12441248)

address the question of how one could integrate the engineered ge-

netic circuits with the rest of the cell environment to attain more

robust functioningthe strict modularity of the Lego-bricks isclearly left behind. On the other hand Ruder, Lu, and Collins

(2011, pp. 12481252)argue that while synthetic biology initially

arose from the combined efforts and insights of a small band of

engineers, physicists, and computer scientists . . . for the field to

reach its full clinical potential, it must become better integrated

with clinicians (Ruder et al., 2011, p. 1251). Thus integration seems

to be the key word; the synthetic biologists in question aim for a

more integrated approach on both the systems and the disciplinary

levels.

1369-8486/$ - see front matter 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.shpsc.2013.03.011

E-mail addresses:[email protected](T. Knuuttila),[email protected](A. Loettgers).1 Accessed 10.08.12.

Studies in History and Philosophy of Biological and Biomedical Sciences 44 (2013) 158169

Contents lists available at SciVerse ScienceDirect

Studies in History and Philosophy of Biological andBiomedical Sciences

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / s h p s c
http://dx.doi.org/10.1016/j.shpsc.2013.03.011mailto:[email protected]:[email protected]://biobricks.org/http://dx.doi.org/10.1016/j.shpsc.2013.03.011http://www.sciencedirect.com/science/journal/13698486http://www.elsevier.com/locate/shpschttp://www.elsevier.com/locate/shpschttp://www.sciencedirect.com/science/journal/13698486http://dx.doi.org/10.1016/j.shpsc.2013.03.011http://biobricks.org/mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.shpsc.2013.03.011


2/12

While these synthetic biologists are calling for more integration

OMalley and Soyer (2012)have recently argued that systems and

synthetic biology already display a wide variety of integrative prac-

tices. As such, systems and synthetic biology may provide an

exemplary case of the various kinds of integration taking place in

contemporary science. In our paper we will concentrate on one

particular type of integration discussed by OMalley and Soyer:

methodological integration. It involves directing a range of meth-

ods at a particular biological system or research problem in order

to gain multidimensional understanding of how the system works

(OMalley & Soyer, 2012, p. 60). Such methodological integration is

characteristic ofcombinational modelingwhereby synthetic biolo-

gists combine the experimentation on model organisms and math-

ematical modeling/simulation with a new type of a model: a

synthetic model. Clearly, this kind of modeling practice involvesthe attempt to gain more multidimensional understanding as it

combines three different methods, all of which are able to deliver

different kinds of knowledge, and all of which are partial and in

need of triangulation with other kinds of evidence. Thus the con-

cept of integration seems to referas the mere word already sug-

geststo the convergence of more complete and uniform

explanations of phenomena often via repeated rounds of piecemeal

adjustment and re-engineering (Chang, 2004; OMalley, 2011).

Such development captures without doubt a great deal of what is

happening in systems and synthetic biologyand in science, more

generally. However, in this paper we wish to also emphasize an-

other aspect of combinational modeling that points towards its

reflexive character.

Combinational modeling may not just lead to greater integra-tion but also to the questioning of some central assumptions and

notions on which synthetic biology is based. This reflexivity is per-

haps most vividly exhibited by the practice of synthetic modeling

and the way in which synthetic models are used to explore the

appropriateness of some features of mathematical models and

their theoretical grounding. In a relatively short period of time,

work on synthetic models has made synthetic biologists aware of

some fundamental differences between human engineered arti-

facts and biological systems. In what follows we will discuss

through our example casesthe insights provided by engineering

synthetic models; that is to say engineering synthetic networks

constructed from genetic material. Before going into the cases we

will briefly discuss the different conceptions of engineering in syn-

thetic biology and describe some general features of synthetic

modeling. Our study is partly based on an empirical study on the

Elowitz lab, which is one of the leading synthetic biology

laboratories.2

2. Engineered artifacts vs. biological systems

2.1. Engineering as the study of functions and system dynamics

In synthetic biology one can distinguish two main approaches:

an application-oriented approach and a basic-science approach.

The application-oriented approach, which aims for instance to de-

sign novel biological parts or organisms for the production of vac-

cines (Ro et al., 2006), biofuels (Bond-Watts, Bellerose, & Chang,

2011), and cancer-killing bacteria (Anderson, Clarke, Arkin, & Voi-

gt, 2006), is often construed as comprising the whole field of syn-

thetic biology. Less visible than this application-oriented approach

is the basic-science approach, which uses synthetic biology and

especially synthetically designed biological parts as a tool for

investigating the basic design principles of gene regulatory net-

works (e.g., Elowitz & Leibler, 2000; Gardner, Cantor, & Collins,

2000). When this line of research took its first steps, one of the

main requirements was to reduce the complexity of biological sys-

tems. The reason for this strategy was not necessarily due to the

reductive vision of the scientists in question but rather their aim

of studying some aspects of biological organization in isolation of

other aspects. This was deemed indispensable for the purposes of

exploring various possible design principleslater to be called

motifs (see below). Another important motivation for this ap-

proach, we suggest, was to explore the concepts, methods, and

techniques imported to systems and synthetic biology from other

disciplines, notably from engineering and physics. This is precisely

the reflexive dimension of the material practice of synthetic

biology.

On the level of modeling methods, the basic-science approach

has been heavily physics-influenced, although many of the central

concepts come from engineering. This raises the question of what

triggered this use of engineering concepts. Interestingly, physi-

cists themselves have advocated the use of engineering concepts

in biology. For example, in the mid-1990s there was some discus-

sion about the appropriateness of transferring concepts from

physics to biology exemplified by programmatic articles such as

From molecular to modular cell biology by Hartwell, Hopfield,

Leibler, and Murray (1999).3 All four authors, two of whom are

physicists (John Hopfield and Stanislas Leibler) and the other two

biologists (Leland Hartwell and Andrew Murray), have made

important contributions in their respective fields of research. In this

article the four authors argue for turning away from the prevailingreductionist approaches in molecular biology that reduce biologi-

cal phenomena to the behavior of molecules ( Hartwell et al., 1999,

p. C47). According to the authors these approaches fail to take into

consideration that biology-specific functions cannot be attributed

to one molecule, but that [. . .] most biological functions arise from

the interaction among many components (Hartwell et al., 1999, p.

C47). To describe biological functions, they go on to claim, we

need a vocabulary that contains concepts such as amplification,

adaptation, robustness, insulation, error correction, and coincidence

detection (Hartwell et al., 1999, p. C47).4 Analogies to engineered

Fig. 1. Cover of the September 2011 issue of Science featuring a special section on

synthetic biology.

2 One of the authors spent four years in the Elowitz lab at the California Institute of Technology observing the daily research practice in this lab.3 Of course, this issue is as old as the mathematical approach to biology; for an overview, see e.g., Kingsland (1985).

4 It deserves to be noted thatHartwell et al. (1999)paint a too reductionist picture of molecular biology also largely ignoring the early attempts to apply engineering conceptsto biologyoften side-by-side with concepts adapted from physics (e.g.,Jacob & Monod, 1961).

T. Knuuttila, A. Loettgers / Studies in History and Philosophy of Biological and Biomedical Sciences 44 (2013) 158169 159


3/12

artifacts were considered appropriate as such items are typically

constructed to fulfill a certain functionlike the parts of biological

organisms.

Interestingly, this amalgamation of engineering concepts and

methods of dynamic systems analysis is understood as the engi-

neering approach by many practicing systems biologists. Thus for

instance Wolkenhauer, in discussing what systems biology can

learn from engineering, more or less equates the engineering ap-proach to the systems-theoretic approach that studies what

role components have in thefunctioningof the cell (Wolkenhauer,

2006, p. 19; italics added). The focus of this engineering approach

is not just on the components but also, importantly, on their non-

linear interactions and the temporal dynamics of these interac-

tions. This is quite a different understanding of the engineering

approach than that offered by the construction-metaphor and the

pursuit of BioBricks.5 However, as a result of combining this sys-

tems-theoretic line of research with the design, manipulation, and

exploration of synthetic biological systems, synthetic biologists have

also begun to consider the limits of some central engineering con-

cepts covering also the basic assumptions of the associated mathe-

matical models. In the following sections we discuss how the

practice of combinational modeling has led scientists to discover

important differences between the control mechanisms of biological

and engineered things.

2.2. Providing control in engineered and biological systems

Control is of central importance in engineered as well as in bio-

logical systems.

However, already early on it was discovered that there are fun-

damental differences between controlling the living behavior of

biological systems and that of engineered artificial systems. Engi-

neered systems typically rely on autonomous control mechanisms.

A thermostat is a good example. In this case the room temperature

(input) is measured, compared to a reference temperature (out-

put), and in the next step the heater is changed in such a way that

the room temperature is adjusted to the reference temperature.The biological solution is more elegant and makes use of an organ-

isms internal oscillating cycles that interact and harmonize their

behavior by coupled oscillations.6 Biological systems need this cyc-

lic organization, since they are autonomous entities able to maintain,

repair, and build themselves using the matter and energy of their

environments.7 Thus biological systems differ crucially from artifi-

cial, engineered systems when it comes to the role of oscillations

in their functioninga point addressed by Brian Goodwin already

in 1960s.

Goodwin was an early mathematical modeler of feedback

mechanisms in biological systems such as metabolic and gene reg-

ulatory control cycles, showing that such mechanisms allowed for

periodic oscillations. Goodwin contrasted the behavior of biologi-

cal oscillators with engineered control systems writing: Theappearance of such oscillations is very common in feedback control

systems. Engineers call them parasitic oscillations because they

use up a lot of energy. They are usually regarded as undesirable

and the control system is nearly always designed, if possible, to

eliminate them (Goodwin, 1963, p. 5). Thus decades before the

emergence of synthetic biology it was already clear that biological

organisms organize their living behavior differently than the way

the engineered artifacts are designed.

Goodwins model and its extensions have been used as basic

templates for various models of oscillatory behavior, including

the circadian clock, which is one of the most studied gene regula-

tory systems. Instead of one clock it actually consists of a large

orchestra of clocks that, on the basis of oscillations on the molec-

ular level, synchronize the functions of the organs in a biological

organism (see e.g., Bechtel & Abrahamsen, 2010, 2011). Although

in comparison to circadian clocks, humanly engineered control sys-tems such as thermostats appear rather simple, but they are still

thought to have something important in common: both make

use of feedback mechanisms. One of the most basic assumptions

in the modeling of control in biological systems is that they make

use of feedback mechanisms. Such feedback mechanisms and their

dynamic properties are typically modeled by using nonlinear dif-

ferential equations, which give rise to oscillations.8 This reliance

of systems and synthetic biology on the engineering notion of the

feedback system, on the one hand, and the realization that oscilla-

tions play a different role in biological systems than in engineered

artifacts, on the other hand, points to the kind of productive double

bindsynthetic biology has over engineering.

This double bind is seen in the continuous dialectic between the

positive and negative analogies drawn between engineered arti-

facts and biological organisms, as the case of feedback systems

and oscillations already shows. Such dialectic prompts scientists

to retrieve resources from different fields of studyand to devise

new strategies to explore the appropriateness of these analogies.

Synthetic modeling can be seen as one such novel strategy.

Namely, up until recently researchers have been uncertain

whether the kinds of feedback systems depicted by the various

mathematical models proposed are really realizable in biological

systems. There remained the possibility that the well-established

ways of mathematically depicting feedback systems and their

dynamics developed by physicists and systems theorists (see e.g.,

Strogatz, 1994) may not be suitable for accounting for how natu-

rally evolved organisms organize themselves. But with the advent

of synthetic biology and synthetic modeling, it became possible to

demonstrate that biological systems may make use of the kinds offeedback systems earlier studied by mathematical modeling.

Moreover it became possible to demonstrate that these feedback

loops can indeed produce the kind of oscillatory behavior exhibited

by gene regulatory systems. Yet this synthetic strategy simulta-

neously yielded new unexpected findings and surprising insights

and, although partially successful, it also made researchers acutely

aware of the differences between biological and engineered

systems. Such recent topics as noise and integration can be directly

related, we suggest, to the combinational, mathematical-cum-

material, strategy of synthetic biology.

3. TheRepressilatorand the combinational strategy

Some of the first synthetic models were constructed to studyhow gene regulatory systems could bring about circadian oscilla-

tions, i.e. the day and night rhythms of the organisms. In their re-

view article, The Pedestrian Watchmaker: Genetic Clocks from

Engineered Oscillators, Cookson et al. describe the typical con-

struction process of a synthetic model in the following way: First,

genetic wiring diagrams are translated into equations that can be

analyzed. [. . .] Next, tools from applied math and computer science

areused to analyze themodel in order to extract the design criteria

5 SeeKnuuttila & Loettgers (in press)for use of various analogies and metaphors in synthetic biology.6 Christiaan Huygens (16291695) first observed the phenomenon of coupled oscillators. The pendulum clocks he mounted on the same non-rigid wall synchronized their

oscillations.7 To what extent biological organisms gain control over their functioning by self-organization arising from interacting oscillations is an open question. Living systems do also

rely on such decoupled controllers as genes (see Bechtel & Abrahamsen, 2011; Bechtel, 2011, for excellent discussions on the role of different oscillations in biological systems).8 Also thermostats produce temperature oscillations but they are designed so as to keep the temperatures close to the target.

160 T. Knuuttila, A. Loettgers/ Studies in History and Philosophy of Biological and Biomedical Sciences 44 (2013) 158169


4/12

for a desired output. Then, modern recombinant DNA techniques

are used to construct gene-regulatory networks in living cells

according to the design specifications. [. . .] Lastly, micro- and

nano-technologies are developed to acquire the precise single-cell

measurements that are needed for comparison with model predic-

tions and design refinement (Cookson, Tsimring, & Hasty, 2009, p.

3931). What seems already clear in the light of this brief character-

ization is that apart from being constructed on the basis of mathe-

matical models, synthetic models are supposed to be used in

combinationwith them.Sprinzak and Elowitz (2005)call the syn-

thetic paradigm a combinational use of experimentation on model

organisms, mathematical modeling, and synthetic modeling.

This combinational modeling is one of the defining strategies of

the basic-science approach of synthetic biology. The basic idea of

the combinational modeling strategy is shown inFig. 2, which is

taken from a review article on synthetic biology by Sprinzak and

Elowitz (2005). As the upper part (a) of the diagram suggests, in

combinational modeling the results gained from the three different

epistemic activities inform each other.

Why do researchers make use of such a combinational model-

ing strategy in studying the organizational principles in biology?A clue can be found from the lower part (b) of the diagram. The

left-hand side of the diagram depicts our present understanding

of the natural gene regulatory circuit of the circadian clock of

Drosophila (fruit fly) consisting of interacting genes and proteins.

The right-hand side in turn depicts a synthetic model of the circa-

dian clock, theRepressilator, which is an oscillatory genetic circuit

introduced in 2000 by Elowitz and Leibler (2000).9 The diagram

indicates the two main differences between the natural and the syn-

thetic system:

1. The natural system exhibits a much higher degree of complex-

ity than the synthetic system.

2. The synthetic circuit has been designed by using different genes

and proteins.

Consequently, synthetic models have the advantage of being

less complex than model organisms. On the other hand, in contrast

to mathematical models they are of the same materiality as biolog-

ical systems (although theRepressilatorwas constructed from dif-

ferent genetic material than the naturally occurring circadian

clocks, a point to which we will return below). This fact of being

of the same materiality as natural systems is crucial for the episte-

mic value of synthetic modeling, since synthetic models are ex-

pected to work in the same way as biological systems. Due to

their right kind of materiality combined with their tightly con-

strained nature, synthetic models have led researchers to discover

such features of the functioning of biological systems that were not

anticipated by mathematical modeling or experimentation with

model organisms.10

3.1. From circuit diagrams to mathematical models

The first step in constructing the Repressilator consisted in

designing a mathematical model used to explore the known basic

biochemical parameters and their interactions through computer

simulations. Synthetic biologists often refer to their mathematical

model as the blueprint for the design and construction of the

synthetic model. What exactly do they mean by that? Is it possible

to say more about the construction of the mathematical model and

how it relates to the synthetic model? Clearly, what is at stake can-

Fig. 2. Combinational modeling according to Sprinzak and Elowitz (2005). The upper part of the diagram depicts the combinational modeling strategy. The lower part

compares a natural gene regulatory network and a synthetic one.

9 TheRepressilatorand the genetic toggle switch (Gardner et al., 2000) were the two first synthetic models. They were published in the same issue of Nature independently of

each other.10 SeeNersessian & Chandrasekharan, (2009) for another discussion on analogical reasoning and combinational modeling involving a physical model, its computational model,

and an intermediary, hybrid computational model. Green (this volume) also contains an excellent discussion of the use of multiple models focusing on the work of Uri Alon andhis group.



5/12

not simply be called translation due to the manifold of construc-

tion decisions and assumptions involved. To get a more informa-

tive account of the transition from a mathematical model to a

synthetic model one needs to answer several questions. How was

the mathematical model arrived at? How is the mathematical

model used to inform the construction of the synthetic model?

And what is the contribution of the other tools, including also tech-

nological artifacts and methods, for the specific construction of a

synthetic model? Here we try shedding some more light on these

questions by first taking a closer look at the design of the mathe-

matical model underlying the Repressilator.

In designing the mathematical model Elowitz and Leibler made

use of the body of theoretical work on regulation mechanisms and

their dynamics in theoretical biology. The systems biologist Uri

Alon calls such simple regulation mechanisms network motifs.They can be depicted by simple diagrams as shown inFig. 3. The

diagram on the left-hand side shows a negative feedback loop

and on the right-hand side a positive feedback loop.

The challenge consists in giving those diagrams an appropriate

mathematical form. This challenge is twofold: Firstly, such seem-

ingly simple mechanisms are represented mathematically by

nonlinear, coupled differential equations leading to complex

dynamics. Secondly, the attempt to give them more biological con-

tent is far from any straightforward procedure. On the one hand,

although there is a large amount of known biochemical details,

most of that knowledge cannot be included in the model for trac-

tability reasons. On the other hand, many crucial mechanisms and

details still remain unknown and the data is often of a wrong kind.

For the kind of dynamic modeling approach of systems and syn-thetic biology one would need e.g. quantitative time-course data,

which is hard to come by (see e.g., Wolkenhauer, 2007).

What the researchers in the basic-science branch of synthetic

biology are after here are the general design principlesof gene reg-

ulatory networks. The network motifs of Uri Alon are made up of

such design principles. Alon characterizes network motifs as

recurring circuits of interactions from which the network is built

(Alon, 2007, p. 542). As Alon describes, such network motifs were

first systematically defined in Escherichia coli, in which they were

detected as patterns that occurred in the transcription network

much more often than would be expected in random networks

(ibid.). A prominent network motif that occurs in about half of

the repressors inE. coli is negative autoregulation, or as it is also

called, a negative feedback loop.A central point in designing a mathematical model of gene reg-

ulation for example consists in translating simple diagrams, as

shown inFig. 3, to a mathematical model. Regulation via negative

or positive feedback are of such a general character that various

mathematical models can be designed to describe these mecha-

nisms and study their dynamics. A good example of this line of

work is provided by the theoretical biologists Ren Thomas and

Richard dAri, who have been studying general types of regulation

mechanisms describing their architecture, interaction, and dynam-

ics. Their book Biological Feedback (1990)11 develops a formal

methodology for analyzing dynamic systems (see also Thomas,

1998). Although directed primarily to biologists the models

presented can be used in other areas of inquiry. As such these

mathematical models of possible regulation mechanisms provide

an important resource for modelers, offering a good example of what

has been called computational templates by Humphreys (2004; see

alsoKnuuttila & Loettgers, 2012). Computational templates are gen-

uinely cross-disciplinary computational devices such as sets of equa-

tions, functions, and computational methods, which can be applied

to different problems in various disciplines. These formal templates

are used by adjusting them to the specific process under investiga-tion; in the case of gene regulation they are adjusted and enriched

by providing more biochemical details. In the following we are going

to examine this process in more detail.

In a simple regulation diagram, of which Fig. 3 provides the

most schematic example, the details of the mechanism are still

invisible. In constructing a mathematical model a first step consists

of describing some basic parts and relations of the mechanism. In

its simplest form such description could be as follows: Firstly a

gene becomes transcribed resulting in the production of mRNA,

from which a protein is produced. In the next step the protein

binds to its own transcription site and by doing so represses the

transcription of its own gene. This leads to a decrease of mRNA

and protein concentration. When the concentration becomes very

low there are not enough proteins to bind at the transcription site

to repress their own production. The production of mRNA and pro-

teins will start again because an increasing number of transcription

sites will not be occupied by the protein and the transcription of

the gene can take place again. This mechanism will lead to oscilla-

tions in the concentration of the mRNA and protein.

How is this process rendered into a mathematical model? A

very common way of describing regulatory interactions makes

use of differential equations. In such a model the concentrations

of mRNA and proteins at a given time tare described by a contin-

ous variable xi(t). The regulatory interactions are represented by

kinetic equations introduced from chemical kinetics.

dxidt fix; 1 i n;

wherefi(x) is the rate law.The rate of change of variable xiis a functionfiof the other con-

centration variablesx= [x1,. . .xn]. The functionfican be of different

forms, depending on how the rate of change is related to the

concentration.

In the case of negative feedback the concentration of mRNA and

proteins is described by x1andx2. The associated regulatory inter-

actions are given by the following set of coupled differential

equations:

dx1dt j1fx2 c1x1

dx2

dt

j2x1 c2x2;

with:

j1;j2= production rate constantsc1;c2= degradation rate constants

The rate law is given by the so-called Hill function:

fx2 H

n

Hn xn

2

;H > 0:

with:

H = genes repression coefficient

x2= protein concentration

n= Hill coefficient.

Fig. 3. Two network motifs: b depicts a negative feedback loop and c a positive

feedback loop (Alon, 2007, p. 451).

11 Thomas and DAris book inspired the construction of the Repressilator; see the end of Section 3.1.



6/12

The Hill function is used in biochemistry for describing thebinding of a ligand12 to a macromolecule. In the case of gene regu-

lation the production rate of the protein is related to the protein con-

centration in such a way that the maximal production rate is reached

when the promoter site is unbound. Half-maximal repression is

reached when the repressor activity is equal to H.

Coming back to the case of theRepressilatorthe negative feed-

back between the three genes is described by a set of six coupled

differential equations.

dmidt mi

a1pnj

a0

dpidt bpi mi

with i lacl; tetR; clj cl; lacl; tetR

In this set of differential equationspiis the concentration of the

proteins suppressing the function of the neighbor genes and mi(where i is lacI, tetR, or cl) the corresponding concentration of

mRNA. The parameter a0 describes what is called leakiness. Evenin the presence of a saturating amount of proteins suppressing

their own production, a slight increase in mRNA can be observed.

a+ a0 describes the maximal production rate in the absence ofany repressors and nis the Hill coefficient. The form of the coupled

set of six differential equations looks different from the general

form introduced before. But it has been already adjusted to the

case of three genes repressing each other via negative feedback.

A further important step in adjusting the general form to thecontext of biology consisted in finding values for parameters such

as the Hill coefficientn as well asa anda0. This has been done byperforming computer simulations and exploring the kinds of ef-

fects a change of those parameters has on the solution of the differ-

ential equations. Elowitz and Leibler did show that there are two

possible types of solutions: The system may converge toward a

stable steady state, or the steady state may become unstable, lead-

ing to sustained limit-cycle oscillations (Elowitz & Leibler, 2000, p.

336). The solutions are depicted inFig. 4. Furthermore, the numer-

ical analysis of the model gave insights into the experimental

parameters showing that [...] oscillations are favoured by strong

promoters coupled to efficient ribosome binding sites, tight

transcriptional repression (low leakiness), cooperative repression

characteristics, and comparable protein and mRNA decay rates

(ibid., p. 336).

In sum, in designing and analyzing their mathematical model

Elowitz and Leibler had at their disposal the research on mathe-matical modeling of regulatory systems. Apart from Thomas and

dAris Biological Feedback (1990), they refer to Winfree (1990)

andGoldbeter (1996). The dynamics of such regulatory systems

can be quite complex and difficult to analyze because of the non-

linear character of the differential equationscaused by the fact

that the rate of the controlled process is a nonlinear function of

the concentration of the regulator.

3.2. The Repressilator and the functional meaning of noise

The structure of theRepressilatoris depicted in theFig. 5.

In the diagram the synthetic genetic regulatory network, the

Repressilator, is shown on the left-hand side and it consists of

two parts. The outer part is an illustration of the plasmid con-structed by Elowitz and Leibler. The plasmid is an extra-chromo-

somal DNA molecule integrating the three genes of the

Repressilator. Plasmids occur naturally in bacteria. In the state of

competence, bacteria are able to take up extra chromosomal DNA

from the environment. In the case of theRepressilatorthis property

allowed the integration of the specific designed plasmid intoE. coli

bacteria. The inner part of the illustration represents the dynamics

between the three genes, TetR, LacI and kcl. The three genes are

connected by a negative feedback loop. The left-hand side of the

diagram shows the Reporterconsisting of a gene producing green

fluorescent protein (GFP), which is fused to one of the three genes

of theRepressilator. The GFP oscillations in the protein level made

visible the behavior of transformed cells allowing researchers to

study them over time by using fluorescence microscopy.The construction of the Repressilatorwas enabled by the devel-

opment of newmethods andtechnologies, such as the construction

of plasmids and Polymerase Chain Reactions (PCR). It is important

to note that the components of the Repressilator(and their number)

had to be chosen in view of what would be optimal for the behav-

ior under study.13 This means that such networks need not be part

of any naturally occurring system. For example the genes used in the

Repressilator do not occur in such a combination in any biological

system, but were chosen and tuned on the basis of the simulations

of the underlying mathematical model and other background knowl-

edge in such a way that the resulting mechanism would allow for

(stable) oscillations.

Fig. 4. Parameter space of the coupled differential equations of the Repressilator

(Elowitz & Leibler, 2000, p. 336).

Fig. 5. The main components of the Repressilator(left-hand side) and the Reporter

(right-hand side) (Elowitz & Leibler, 2000, p. 336).

12

A ligand is a signal-triggering molecule.13 In the case of the Repressilator the order in which the genes are connected to each other turned out to be crucial as well.



7/12

What, then, was the point of constructing the Repressilatorif it

clearly was an engineered synthetic network with no counterpart

in naturally evolved genetic networks? The answer we have al-

ready hinted at was that it was an epistemic tool designed for

answering certain kinds of theoretical questions (Knuuttila, 2011,

see below). One question addressed by the Repressilatorwas due

to the fact that the model templates, methods, and concepts used

in systems and synthetic biology were not originally devised withbiological organisms in mind. Another question concerned the ab-

stract and highly idealized nature of these models and their under-

determination by the available data (see Knuuttila & Loettgers,

2011). As such, mathematical models, even combined with exper-

imental evidence, were unable to settle the question of whether

the various network designs proposed, for example in the context

of circadian clock research, could actually work in biological organ-

ismseven though they were able to create the sought-after oscil-

lations. Moreover by implementing the synthetic genetic network

into a cell it is exposed to some further constraints of natural bio-

logical systems, thus providing insight into the modularity of the

circadian mechanisms (see below).

In fact the Repressilatorsparked a new line of research as a di-

rect result of its limited success. In contrast to the mathematical

model underlying it the Repressilatordid not show the expected

behavior: regular oscillations. Instead the oscillations turned out

to be noisy. As a result Elowitz and Leibler constructed a stochastic

version of the original deterministic model. Computer simulations

of this model suggested that stochastic fluctuations in gene expres-

sion could be the cause of this noisy behavior. Spudich and Kosh-

land (1976) had already suggested that stochastic fluctuations

could be due to the low number of molecules in cells. However,

at that time no means existed for the direct observation of such

fluctuations on a molecular level. This only became possible with

the introduction of Green Fluorescent Proteins and synthetic mod-

eling, which contributed importantly to the emergence of a new

research program examining the sources and role of noise in bio-

logical systems (e.g.,Elowitz, Levine, Siggia, & Swain, 2002; Swain,

Elowitz, & Siggia, 2002; see also Loettgers, 2009). As a result of thisnew line of research noise has also gained a functional meaning:

Biological systems appear to make good use of noise in diverse pro-

cesses, including development (Neildez-Nguyen et al., 2008), dif-

ferentiation (e.g., genetic competence; see Cagatay et al., 2009),

and evolution (Eldar & Elowitz, 2010). This contrasts with engi-

neering where noise is usually considered as a disturbance. In thus

allowing noise also a functional meaning, this new research pro-

gram actually draws a further negative analogy to engineered con-

trol systems (apart from the different role of oscillations in

engineered artifacts and natural systems, pointed out already by

Goodwin, see above). And the studies on noise point towards still

another difference. Namely, apart from internal noise, there re-

mained the possibility that the noisy behavior could also have been

caused by external noise coming from the cell environment. TheRepressilatorwas probably not so modular as it was supposed to

be, that is, it did not form as isolated a module in its host system

as was expected. Indeed, apart from noise, modularity is another

engineering concept whose limits have been questioned by recent

research in synthetic biology. This line of research has made syn-

thetic biologists more concerned with the integration of synthetic

parts with the rest of the cell.

4. From modularity to integration

4.1. Investigating the modularity assumption

Modular organization is among the most basic and important

assumptions of synthetic biology but also one of the most con-tested. Since its beginning synthetic biology has had a complex

relationship to the assumption of modularity: on the one hand

synthetic biology relies on the assumption of modular organization

in view of its aim to design autonomous modules of interacting

components that would give rise to a specific function/behavior.

On the other hand each synthetic biological system also functions

as a test to which extent the assumption of the modular organiza-

tion is justified. In interpreting such experiments one should keep

in mind, though, a distinction between the modular organization inbiological systems, which are outcomes of a long evolutionary pro-

cess, and the modular organization of synthetic systems, which

have been designed in the lab. The modular design of synthetic

systems does not necessarily presuppose that natural biological

systems are organized in a modular way. But it has been generally

supposed that for the synthetic modular strategy to succeed,

biological systems have to allow for the integration of synthetic

systems in a modular fashion. As in the case of noise, if non-

modular organization should turn out to be an advantage in realiz-

ing specific functions in synthetic systems, it would be a good

strategy to make use of it. Looking at recent developments in

synthetic biology there are indeed attempts to make use of more

integrated structures. However, as in the case of noise, one has to

pay the price of losing partial control over the synthetic system.

4.2. Dual-feedback oscillator

Another good example of how synthetic modeling can lead to

unexpected findings is provided by the work Jeff Hasty and his

co-workers. They designed a dual-feedback oscillator which, by

connecting more tightly with the host cell, shows surprisingly ro-

bust behavior. As to the number of genes, the design of the dual-

feedback oscillator is even simpler than that of the Repressilator.

It consists of only two genes: an activator and a repressor. The

expression of either gene can be enhanced by the activator protein

and blocked by the repressor protein. Both proteins function as

transcription factors for both genes. As such in this synthetic oscil-

lator two motifs, a positive and a negative feedback, are com-

bined: a promoter drives the production of both its own activatorand repressor (seeStricker at al., 2008). In a later review article

the group expanded the earlier discussion of the synthetic oscilla-

tor (Cookson et al., 2009). In this latter article they introduce the

synthetic oscillator by explaining how it was based on both exper-

imental work on the Drosophila melanogaster(fruit fly) clock and

earlier theoretical work, i.e. a mathematical model set forth by

Hasty, Dolnik, Rottschfer, and Collins (2002) which, as they put

it, is theoretically capable of exhibiting periodic behavior (Cook-

son et al., 2009, p. 3932).Fig. 6shows the network diagram of the

dual-feedback oscillator.

The synthetic network consists of two genes: araC and lacI. A

hybrid promoter plac/ara1 (the two small adjacent boxes) drives

transcription ofaraCand lacI, forming positive and negative feed-

back loops. It is activated by the AraC protein in the presence ofarabinose and repressed by the LacI protein in the absence of

Fig. 6. A diagrammatic representation of the dual-feedback oscillator (Strickeret al., 2008, p. 516).



8/12

isopropyl b-D-1thiogalactopyranoside (IPTG). As in the case of the

Repressilatorthe oscillations are made visible by a green fluores-

cent protein (GFP). This synthetic system provides a material sys-

tem allowing for the study of the oscillatory dynamic of the

dual-feedback system: the basic properties and conditions of the

oscillatory dynamic can be estimated on the basis of the compo-

nents of the model and their interactions. By adding arabinose

and IPTG to the system the promoter becomes activated and thesystems two genes, araC and lacI, become transcribed. An in-

creased production of the protein AraC in the presence of arabinose

results in a positive feedback loop that increases the promoter

activity. On the other hand, the increase in LacI production results

in a linked negative feedback loop that decreases promoter activ-

ity. The difference in the promoter activities of the two feedback

loops leads to the oscillatory behavior.

On the basis of the analysis of the synthetic system the

researchers made several observations that were both unexpected

and difficult to reconcile with the original mathematical model

that nevertheless formed the basis of the synthetic model. The

problem was, quite surprisingly, that the oscillator performed al-

most too well as Nandagopal and Elowitz put it (2011, p. 1244).

Two of the observations proved particularly important in this re-

spect. First, the time delays inherent in the process of gene expres-

sion that were initially ignored were shown to be critical for the

robust oscillations (Stricker et al., 2008). They were first over-

looked because they are shorter than the overall period of the oscil-

lator. Second, perhaps the most drastic of the surprising findings is

the relationship of the parameter set and the robustness of the

oscillations. While the mathematical model predicted robust oscil-

lations only for a small set of parameters, the synthetic model con-

tradicted the mathematical model by showing robust oscillations

for a much larger set of parameter values. The researchers were

perplexed by the fact that [i]t was difficult to find inducer levels

at which the system did not oscillate! (Cookson et al., 2009, p.

3934; italics in the original).

These contradictory results led the researchers in question to

reconsider the original model. Cookson et al. wrote about this asfollows: In other words, it became increasingly clear that the ob-

served oscillations did not necessarily validate the model, even

though the model predicted oscillations. We were able to resolve

the discrepancy between the model and theory by reevaluating

the assumptions that led to the derivation of the model equations

(Cookson et al., 2009, p. 3934). The key finding in the reevaluation

of the mathematical model was that a component in the complex

biochemical process that had been underestimated and left out

from the mathematical model turned out to be of central impor-

tance for the robustness of the observed oscillations in the syn-

thetic model. Without going into greater detail, the processes

involved in the production of the transcription factors led to a post-

translational coupling between the activator (AraC) and the repres-

sor (LacI), which is a consequence ofan unintended interaction withthe host cell.

These unintended interactions with the host cell show that syn-

thetic systems need not comprise perfect modules; instead inter-

actions with the cell environment may occur and can even be

advantageous, as the case of the dual-feedback oscillator shows.

In fact there is a new trend in synthetic biology targeting the ques-

tion of how synthetic systems could be designed in such a way that

they are more closely integrated with the endogenous cellular pro-

cesses of the host organisms (e.g.,Nandagopal & Elowitz, 2011, see

below). Hence somewhat paradoxically, although the program of

synthetic biology was originally based on the idea of constructing

isolated synthetic circuits from well-characterized components, it

seems actually to accumulate evidence on how loosening the

assumption of modularity can lead to new and unexpected possi-

bilities. Following this route means the development of new strat-

egies for designing and probing synthetic systems. A first step in

this direction is taken byNandagopal and Elowitz (2011)who, in

a recent article, discuss the various strategies for integration.14

4.2.1. Strategies for integrationThe recent research on noise and integration in the field of syn-

thetic biology points toward new challenges and the need for novel

strategies. In regard to the question of noise, theuse ofnoise in the

design and engineering of synthetic systems could become critical.

Likewise, when it comes to the (non)modular organization, the

question becomes one of how to integrate the components of syn-

thetic systems with those of the host cell to support the perfor-

mance of the synthetic system. Nandagopal and Elowitz (2011)

put forward one possible strategy. The two authors explicate what

they mean by integration on the systems level by asking: Does a

synthetic circuit need to operate independently of its host to func-

tion reliably? (Nandagopal & Elowitz, 2011, p. 1244) and go on to

discuss the work of Jeff Hasty and his co-workers on a simple dual-

feedback oscillator (Section 3.2, see above). What is especially

interesting about Nandagopal and Elowitzs discussion is that they

make explicit the importance of the cellular milieu for the func-

tioning of the synthetic gene circuits. NeitherStricker et al. (2008)

norCookson et al. (2009)spell that out, although it is implicit in

their findings.Stricker et al. (2008)show that time delays inherent

to the process of gene expression were critical for the robust func-

tioning of the oscillator. In addition Cookson et al. (2009) discussed

unintended interactions with the host cell components that im-

proved the precision of the oscillator. Yet neither Stricker et al.

(2008)norCookson et al. (2009)discuss the consequences of their

findings regarding the basic assumption of synthetic biology that

synthetic systems can be integrated into a host cell without inter-

fering with/or disrupting each others functions.

To be sure, the question of modularity is a tricky one. Synthetic

biology creates circuits that would not survive in the naturalworld. Such a synthetic system makes use of specific processes

provided by the host cell, which are needed to keep the synthetic

system alive. Thus the synthetic system depends in a vital fashion

on the host cell but despite this dependency the two systems can

be claimed to remain independent on the level of network dynam-

ics, and on the level of the functions arising from the dynamic

interaction between the components of the network. Such interre-

lationship between a host cell and a synthetic system has been

considered a necessary requirement for retaining control over the

synthetic system and its function once it has become part of a host

cell. This central assumption of synthetic biology is based on the

more general assumption of modular organization of biological

systems. Nandagopal and Elowitz give the following description

of the relationship between synthetic systems and host cells:The view was that underlying cellular processes could be used

to support the synthetic circuits, for example, by providing the

gene expression machinery, but that the two layers could function

independently (Nandagopal & Elowitz, 2011, p. 1244).

What do Nandagopal and Elowitz mean by this? In order to

exemplify the situation one can draw an analogy to computer pro-

grams. Many computer programs make extensive use of standard

routines for calculating specific mathematical functions. These

routines are called upon when needed in the course of the pro-

gram. They are programmed and connected to the computer pro-

gram in such a way that the interaction is restricted to calculate

the function from a given input and to deliver the result. But the

14 This article appeared in the September 2011 issue ofScience, see above.

http://-/?-http://-/?-


9/12

program will not change the routine it calls upon, nor the other

way around.The crucial question is what happens with the assumption of

the respective independent functioning of the synthetic system

and the host cell in the light of the findings of Hasty and his co-

workers? Do these findings give reason enough to announce the

breaking down of the modular relationship between a synthetic

and a biological systemor, to go one step further, to question

the assumption of a modular organization of biological systems

in general? When discussing the possible consequences of these

findings, Michael Elowitz stated:

I think its hard to say in general what the implications of these

kinds of interactions are. One way to look at them is as prob-

lems that need to be eliminated through good design. Another

way is as opportunities to implement non-canonical designs.

A final possibility is that these are coupling mechanisms that

are actually used physiologically at least in some cases. Its

interesting in this light that in competence there is a somewhat

analogous mechanism whereby the protease adapter MecA

mediates an effective interaction between two of its substrates,

ComS and ComK.15

It is interesting that synthetic biologists, like Nandagopal and Elo-

witz, do not rush to question the modularity assumption as such

but instead invoke the idea of integration and the possible ways

of dealing with host cell interactions.16 If there is a coupling be-

tween a synthetic system and a host cell, then there are two possible

ways of handling them. Either one has to come up with novel de-

signs that avoid such couplings or, if they turn out to be advanta-

geous, one needs to look for ways to integrate them into thedesign of synthetic systems. Nandagopal and Elowitz are clearly opt-

ing for a second route. They call for synthetic systems that integrate

more closely with endogenous cellular processes (Nandagopal &

Elowitz, 2011, p. 1244). With this step, they suggest, the field would

move away from its original aim of designing autonomous genetic

circuits that could function as independently as possible from

endogenous cellular circuits or even functionally replace endoge-

nous circuits (ibid., p. 1244).

Nandagopal and Elowitz use a tri-partite picture (Fig. 7) to de-

pict what they think will be one of the big changes in the practice

of synthetic biology: Future progress will require work across a

range of synthetic levels, from rewiring to building autonomousand integrated circuits de novo (ibid., p. 1244).

In the diagram depicted inFig. 7Elowitz and Nandagopal intro-

duce what they call the continuum of synthetic biology. In this

continuum one moves from the wild type towards fully autono-

mous synthetic systems increasing the degree of the synthetic part

of the system. How is this increase in the synthetic part achieved?

There are several options. One can follow the traditional ap-

proach of designing an assumedly modular genetic circuit and

introducing it into the wild type. As the example ofStricker et al.

(2008)andCookson et al. (2009)nevertheless showed, unintended

interactions can occur (gray arrows) that could be difficult to con-

trol. An alternative approach, propagated by Nandagopal and Elo-

witz, consists in first rewiring the genetic circuit in the wild type

and then, in a second step, implementing a synthetic circuit intothe rewired circuit. This rewiring of the existing genetic circuits of-

fers, firstly, a way to explore the design principles on which the ge-

netic circuit is based and, secondly, a possibility of using these

insights to avoid unintended interactions with the host cell. Inter-

estingly, as has been shown in a number of studies in which the

strategy of rewiring has been used, the actual biological design

principles are often counter-intuitive (see for example agatay,

Turcotte, Elowitz, Garcia-Ojalvo, & Sel, 2009). Nature appears to

have used solutions that differ from those of engineers.

As a consequence of the rewiring strategy the resulting engi-

neered circuit is only partially independent. However, for the engi-

neering purposes, as high a modularity as possible is usually

sought because of its controllability. In order, then, to get an inde-

pendent circuit that would be based on the insights gained fromthe exploration of the rewired circuit, one would integrate the

function of the rewired circuit design into an autonomous genetic

circuit. This strategy allows for suppressing unwanted interactions

with the host cell but also implementing interactions, which sup-

port the function in question. In more general terms the proposed

strategy tries to balance the need for control and the possibility of

taking advantage of the interactions with the host cell. In such a

way the engineering of synthetic systems becomes increasingly in-

spired by biological systemsa point that has recently been

stressed by several synthetic biology research programs.17

Fig. 7. The continuum of synthetic biology (Nandagopal & Elowitz, 2011, p. 1244).

15 Personal communication by email.16 In an interview, Elowitz also discussed the fact that the cells in which the first generation synthetic models were implanted tended to grow more slowly, which may also

point towards some interactions taking place.17 See e.g., http://wyss.harvard.edu/viewpage/264/a-new-model(Accessed 5 January 2012).

http://wyss.harvard.edu/viewpage/264/a-new-modelhttp://wyss.harvard.edu/viewpage/264/a-new-model


10/12

In sum, there are different ways in which one can deal with host

cell interactions. The findings ofStricker et al. (2008)andCookson

et al. (2009)can also prompt researchers to perceive natural sys-

tems in novel ways and actively search for similar couplings and

mechanisms in biological systems as suggested by the behavior

of the synthetic models. This could lead to some unexpected find-

ings regarding the organization of biological systems. Thus the dis-

cussion on host cell interactions shares some importantsimilarities with the case of noise. It seems that in synthetic biol-

ogy the engineering concepts are changing their meanings and

becoming more and more adjusted to biological systemspartly

as a result of the practice of synthetic modeling.

5. Conclusions: Basic science through engineering

Integration is a word that characterizes many things taking

place in systems and synthetic biology, from the level of the con-

struction of synthetic systems to the level of research methodology

and interdisciplinary relationships. The synthetic biologists, whose

work we have studied here, focus on biological integration, that is,

how to integrate synthetic systems more closely with their host

organisms. We have in turn examined combinational modeling,

which is one form of methodological integration characteristic of

synthetic biologyand also utilized by synthetic biologists to study

biological integration (see OMalley & Soyer, 2012).18 Combina-

tional modeling refers to a process whereby synthetic biologists

use experimentation with model organisms, mathematical model-

ing, and synthetic modeling to inform each other. Through combina-

tional modeling synthetic biologists have often gained unexpected or

otherwise surprising insights into the organization of biological

organisms. While there is a long story to be told on how mathemat-

ical modeling and experimentation on model organisms have con-

tributed to each other in circadian-clock research (see e.g., Bechtel,

2010; Bechtel & Abrahamsen, 2010), we have focused here on the

relationship between mathematical modeling (and simulation) and

synthetic modeling.What is the kind of integration that happens in combinational

modeling? Two notions presented in the previous philosophical

literature seem especially relevant: triangulation and iteration.

Triangulation is the use of different epistemic means to study

the same phenomenon. It is often conceptualized as the combi-

national use of independent means to triangulate the existence

and character of a common phenomenon, object or result

(Wimsatt, 2007, p. 43; see also Wimsatt, 1981). There has been

some discussion on how independent these triangulated episte-

mic methods can be (e.g., Orzack & Sober, 1993); because, often,

as in the cases presented here, this is not the case. Quite the

contrary, in combinational modeling, mathematical models and

synthetic models are closely linkedwhile, on the other hand,

the materiality of synthetic models simultaneously brings in anindependent component. It seems that it is precisely this dialec-

tic of relatedness and independence that gives combinational

modeling its epistemic leverage leading to iterative cycles of

modeling.

Epistemic iteration refers to a process of corrective evolu-

tion whereby the step-wise cyclic application of different meth-

ods and new layers of knowledge gradually enhance the

achievement of the epistemic goals sought for (Chang, 2004).

Such iterative process is indeed discernible in the cases we have

discussed. Both in the case of the Repressilator and the dual-

feedback oscillator the behavior of the synthetic model led to

the improvement or modification of the mathematical model

that informed the construction of it. Elowitz and Leibler resorted

to a stochastic version of their original deterministic model to

explain the noisy oscillations of theRepressilator. Hasty and his

co-workers in turn complemented and revised their mathemati-

cal model in view of the unexpected importance of the time

delays and the interactions with the host organism (Stricker

et al., 2008; Cookson et al., 2009). Consequently, these casesconfirm OMalleys observation that again and again in syn-

thetic biology [. . .] false assumptions about the system are

revealed in the construction of devices (OMalley, 2011,

p. 408). But these assumptions being questioned may also point

beyond the iterative cycle, towards new questions, novel re-

search agendas and, importantly, the re-evaluation of the basis

on which synthetic biology is built.

Namely the material-cum-theoretical nature of synthetic

models, such as the Repressilator and the dual-feedback oscilla-

tor, and the possibility of directly manipulating biological com-

ponents and networks offered by them, hasapart from many

valuable insights into biological organizationalso disclosed the

limitations of any single-minded engineering approach. Novel re-

search topics such as noise and integration can be directly re-lated to synthetic modeling and the way it is combined with

mathematical modeling. Consequently, then, it seems partly mis-

guided to claim that synthetic biology aims at construction,

whereas the objective of systems biology is to understand exist-

ing biological systems (Calvert & Fujimura, 2011, p. 160). While

this may be a correct observation as regards even a major part of

synthetic biology, material construction can also be a way to

take the systems and synthetic biology approach a step further,

even theoretically.

The construction of synthetic systems may be epistemically

informative in several ways. Clearly, a lot of knowledge, partly

embedded in the technologies and techniques used is needed to

construct synthetic systems in the first place. A well-functioning

tool validates the theoretical knowledge it embodies and, evenfailures, say, of the various visions of standardization also teach

us something about biological systems (e.g., OMalley, 2011,

p. 409). This seems to be the insight behind the idea of thing

knowledge and the way it is juxtaposed with theoretical knowl-

edge (Baird, 2004; on thing knowledge, see also Gelfert this

volume). But here we have been dealing with a material con-

struction strategy that appears also to go beyond mere thing

knowledge being more directly involved in really exploring the

theoretical assumptions made in modeling the gene regulatory

behavior.

In this respect it is instructive to consider how the synthetic

biologists, whose work we have studied, themselves conceive of

what they are doing.Cookson et al. (2009)consider the construc-

tion of synthetic oscillators as basic science through engineering.

They write:

The possibility of a minimal core network driving robust cellu-

lar behavior has inspired the development of an alternative

approach to the study of gene-regulatory networks: create the

network, beginning with a one or two-component system and

then rebuild the network from the bottom up. In this way, we

can gradually assemble increasingly complex systems that

mimic the native network, while maintaining at each stage

the ability to model and test the network in a tractable experi-

mental system. (Cookson et al., 2009, p. 3933)

18 It would be interesting to study to which extent biological integration (non-modularity, or soft modularity) requires methodological integration. Our hunch is that as

biological integration increases the complexity of biological organization, its study also requires multiple methods due to the fact that each method has its own characteristicconstraints. We wish to thank the anonymous referee for raising this interesting question.



11/12

In an interview Michael Elowitz explained to us his motivation for

engaging in synthetic modeling in the following way:

It seemed like what we really wanted to do is to build these cir-

cuits to see what they really are doing inside the cell . . . that

would be in a way the best test of it. . .of what kind of circuits

were sufficient for a particular function.

From the perspective of modeling what seems especially interest-

ing is the stress Elowitz, Hasty and their co-workers place on

minimality, sufficiency and tractability. In this synthetic models

are like mathematical models that are tightly constrained systems

that, by their design, enable the study of certain pending scien-

tific questions. What is important for this kind of modeling strat-

egy is precisely its minimality and tractability. One seeks to

implement only those kinds of principles and assumptions that

are supposed to be sufficient and/or necessary to produce the

behavior under investigation. The synthetic model can then be

considered as an epistemic tool constructed to explore some basic

assumptions and network architectures used in the study of gene

regulatory systems (see Knuuttila, 2011).

There is also a sense in which synthetic modeling comes close

to experimenting with theory, as discussed in the context of simu-

lation (e.g.,Dowling, 1999; Fox Keller, 2003; Winsberg, 2003). Yet

as opposed to simulation models transformed into a computational

algorithm and run on a digital computer, here the theoretical mod-

el rendered as a synthetic model is of the same natural kind as

the native networks as well as being embedded in a simulation

environment of the same materiality, i.e., the host organism

(on the same materiality see e.g., Morgan, 2003, 2005; Guala,

2002). In being constructed from genetic material and embedded

in a natural cell environment, the synthetic model is supposed to

be subject to the same constraints as the natural gene regulatory

networks. It seems that it is precisely this combination of theoret-

ical and material constraints that make synthetic modeling and

combinational strategy so fruitful in studying some theoretical

questions.To conclude, in this article we have discussed the basic-science

approach to synthetic biology and showed how the purposefully

designed synthetic parts can function as powerful epistemic tools

in studying biological organization. This line of research exhibits

a sophisticated reflexive double bind to engineering: engineering

approaches are made use of in constructing synthetic networks,

but at the same time these engineering principles are also explored

in an attempt to find out whether and to which extent they apply

to biological organisms. The recent concerns on noise and integra-

tion are related to this line of research, which seeks the ways in

which biology could increasingly inform and inspire engineering.

It also shows that some traditional engineering aims, like control-

lability, may need to be compromised or implemented in novel,

and so far largely unforeseen ways, in view of attaining robust bio-logical functioning. The toy-biology of Lego-bricks is thus as far re-

moved from these visions and associated challenges as one can

even imagine.

Acknowledgements

We are grateful to Maureen OMalley for her important com-

ments on this paper. We also thank the participants of the sym-

posium on Philosophical Perspectives on Synthetic Biology,

Helsinki, May 2011, for stimulating discussions on synthetic biol-

ogy that have proved beneficial for this work. We also wish to

thank the Equinox Graphics for permission to use their image

on the cover of the September 2011 issue of Science free of

charge.

References

Alon, U. (2007). An introduction to systems biology: Design principles of biologicalcircuits. Boca Raton, FL: Chapman & Hall/CRC Press.

Anderson, J. C., Clarke, E. J., Arkin, P. A., & Voigt, C. A. (2006). Environmentallycontrolled invasion of cancer cells by engineered bacteria. Journal of MolecularBiology, 355, 619627.

Baird, D. (2004). Thing knowledge. Berkeley & Los Angeles: University of CaliforniaPress.

Bechtel, W. (2010). The downs and ups of mechanistic research: Circadian rhythmresearch as an exemplar. Erkenntnis, 73, 313328.Bechtel, W. (2011). Mechanism and biological explanation.Philosophy of Science, 78,

533557.Bechtel, W., & Abrahamsen, A. (2010). Dynamic mechanistic explanation:

computational modeling of circadian rhythms as an exemplar for cognitivescience.Studies in History and Philosophy of Science, 41, 321333.

Bechtel, W., & Abrahamsen, A. (2011). Complex biological mechanisms: Cyclic,oscillatory, and autonomous. In C. A. Hooker (Ed.). Philosophy of complexsystems. Handbook of the philosophy of science (Vol. 10, pp. 257285). Oxford:Elsevier.

Bond-Watts, B. B., Bellerose, R. J., & Chang, M. C. (2011). Enzyme mechanism as akinetic control element for designing synthetic biofuel pathways. NatureChemical Biology, 7, 222227.

agatay, T., Turcotte, M., Elowitz, M. B., Garcia-Ojalvo, J., & Sel, G. M. (2009).Architecture-dependent noise discriminates functionally analogousdifferentiation circuits. Cell, 139(3), 512522.

Calvert, J., & Fujimura, J. (2011). Calculating life? Duelling discourses ininterdisciplinary systems biology. Studies in History and Philosophy of

Biological and Biomedical Sciences, 42, 155163.Chang, H. (2004). Inventing temperature: Measurement and scientific progress. Oxford:

Oxford University Press.Cookson, N. A., Tsimring, L. S., & Hasty, J. (2009). The pedestrian

watchmaker: Genetic clocks from engineered oscillators. FEBS Letters,583, 39313937.

Dowling, D. (1999). Experimenting on theories. Science in Context, 12, 261273.Eldar, A., & Elowitz, M. B. (2010). Functional roles for noise in genetic circuits.

Nature, 467, 167173.Elowitz, M. B., & Leibler, S. (2000). A synthetic oscillatory network of transcriptional

regulators.Nature, 403(6767), 335358.Elowitz, M. B., Levine, A. J., Siggia, E. D., & Swain, P. S. (2002). Stochastic gene

expression in a single cell. Science, 297, 11831186.Endy, D. (2005). Foundations for engineering biology. Nature, 438, 449453.Fox Keller, E. (2003). Models, simulation and computer experiments. In H. Radder

(Ed.), The philosophy of scientific experimentation (pp. 198215). Pittsburgh:University of Pittsburgh Press.

Gardner, T. S., Cantor, C. R., & Collins, J. J. (2000). Construction of a toggle switch inEscherichia coli.Nature, 403(6767), 339342.

Goldbeter, A. (1996). Biochemical oscillations and cellular rhythms: Themolecular bases of periodic and chaotic behaviour. Cambridge: CambridgeUniv. Press.

Goodwin, B. (1963). Temporal organization in cells. London: Academic Press.Guala, F. (2002). Models, simulations, and experiments. In L. Magnani & N.

Nersessian (Eds.),Model-based reasoning: Science, technology, values(pp. 5974).New York: Kluwer.

Hartwell, H. L., Hopfield, J. J., Leibler, S., & Murray, W. A. (1999). From molecular tomodular cell biology.Nature, 402, C47C52.

Hasty, J., Dolnik, M.,Rottschfer, V.,& Collins, J. J. (2002). Synthetic gene network forentraining and amplifying cellular oscillations. Physical Review Letters, 88,148101148104.

Humphreys, P. (2004). Extending ourselves: Computational science, empiricism, andscientific method. Oxford: Oxford University Press.

Jacob, F., & Monod, J. (1961). Genetic regulatory mechanisms in the synthesis ofproteins.Journal of Molecular Biology, 3, 318356.

Kingsland, S. (1985). Modeling nature. Chicago and London: The University ofChicago Press.

Knuuttila, T. (2011). Modeling and representing: An artefactual approach. Studies inHistory and Philosophy of Science, 42, 262271.Knuuttila, T., & Loettgers, A. (2011). Causal isolation robustness analysis: The

combinatorial strategy in synthetic biology. Biology and Philosophy, 26(5),773791.

Knuuttila, T., & Loettgers, A. (2012). The productive tension: Mechanisms vs.templates in modeling the phenomena. In P. Humphreys & C. Imbert (Eds.),Representations, models, and simulations (pp. 224). Routledge.

Knuuttila, T., & Loettgers, A. (in press). Varieties of noise: Analogical reasoning insynthetic biology.Studies in History and Philosophy of Science.

Loettgers, A. (2009). Synthetic modeling and the emergence of a dual meaning ofnoise. Biological Theory, 4(4), 340356.

Morgan, M. (2003). Experiments without material intervention: Modelexperiments, virtual experiments and virtually experiments. In H. Radder(Ed.), The philosophy of scientific experimentation (pp. 216235). Pittsburgh:University of Pittsburgh Press.

Morgan, M. (2005). Experiments versus models: New phenomena, inference andsurprise.Journal of Economic Methodology, 12, 317329.

Nandagopal, N., & Elowitz, M. B. (2011). Synthetic biology: Integrated gene circuits.

Science, 333, 12441248.



12/12

Neildez-Nguyen, T. M. A., Parisot, A., Vignal, C., Rameau, P., Stockholm, D., Picot, J.,et al. (2008). Epigenetic gene expression noise and phenotypic diversification ofclonal cell populations. Differentiation, 76(1), 3340.

Nersessian, N. J., & Chandrasekharan, S. (2009). Hybrid Analogies in conceptualinnovation in science. Cognitive Systems Research Journal, Special Issue:Integrating Cognitive Abilities, 10, 78188.

OMalley, M. A. (2011). Exploration, iterativity and kludging in synthetic biology.Comptes Rendus Chimie, 14, 406412.

OMalley, M. A., & Soyer, O. S. (2012). The roles of integration in molecular systemsbiology. Studies in History and Philosophy of Biological and Biomedical Sciences, 43,

5868.Orzack, S. H., & Sober, E. (1993). A Critical assessment of Levinss The strategy of

model building in population biology (1966). The Quarterly Review of Biology,68, 533546.

Ro, D. K., Paradise, E., Quellet, M., Fisher, K., Newman, K., Ndgundu, J., Ho, K., Eachus,R.,Ham,T., Kirby,J., Chang, M. C. Y.,Withers, S., Shiba, Y.,Sarpong, R.,& Keasling,

J. (2006). Production of the antimalarial drug precursor artemisinic acid inengineered yeast.Nature, 440, 940943.

Ruder, W. C., Lu, T., & Collins, J. J. (2011). Synthetic biology moving into the clinic.Science, 333, 12481253.

Sprinzak, D., & Elowitz, M. B. (2005). Reconstruction of genetic circuits.Nature,438(7067), 443448.

Spudich, J. L., & Koshland, D. E. (1976). Non-genetic individuality: Chance in thesingle cell.Nature, 262, 467471.

Stricker, J., Cookson, S., Bennet, M. R., Mather, W. H., Tsimring, L. S., & Hasty, J.(2008). A fast, robust and tunable synthetic gene oscillator. Nature, 456,516519.

Strogatz, S. (1994). Nonlinear dynamics and chaos: With applications to physics,biology, chemistry, and engineering. Cambridge, MA: Perseus Books.

Swain, P. S., Elowitz, M., & Siggia, E. D. (2002). Intrinsic and extrinsic contributionsto stochasticity in gene expression. Proceedings of the National Academy ofSciences of the United States of America, 99(20), 1279512800.

Thomas, R. (1998). Lawsfor the dynamics of regulatory circuits. International Journalof Developmental Biology, 42, 479485.

Thomas, R., & DAri, R. (1990). Biological feedback. Boca Raton, FL: CRC Press.Wimsatt, W. C. (2007).Re-engineering philosophy for limited beings: Approximations

to reality. Cambridge: Harvard University Press.Wimsatt, W. C. (1981). Robustness, reliability, and overdetermination. In M. B.

Brewer & B. Collins (Eds.), Scientific inquiry and the social sciencesA volume inhonor of Donald T. Campbell (pp. 124163). San Francisco: Jossey-Bass.

Winfree, A. T. (1990). The geometry of biological time. New York: Springer.Winsberg, E. (2003). Simulated experiments: Methodology for a virtual world.

Philosophy of Science, 70, 105125.Wolkenhauer, O. (2006). Engineering approaches: What can we learn from it in

systems biology? In R. van Driel (Ed.),Systems biology: A grand challenge for EuropeEuropean science foundation forward look report(pp. 1921). Strasbourg: IREG.

Wolkenhauer, O. (2007). Why systems biology is (not) called systems biology?BIOForum Europe, 4(2007), 3839.


basic science through engineering synthetic modeling and the idea of biology inspired engineering...

Documents