Theoretical principles for biology: Variation
Theoretical principles for biology: Variation
Ma¨el Mont´evila,b,∗, Matteo Mossiob , Arnaud Pochevillec , Giuseppe Longod,e
a Laboratoire”Mati`ere et Syst`
emes Complexes” (MSC), UMR 7057 CNRS, Universit´ e Paris 7 Diderot, Paris, France
b Institut
d’Histoire et de Philosophie des Sciences et des Techniques (IHPST) - UMR 8590, 13, rue du Four, 75006 Paris, France
c Department of Philosophy and Charles Perkins Centre, The University of Sydney, Sydney, Australia
d Centre Cavaill`
es, R´epublique des Savoirs, CNRS USR3608, Coll` ege de France et Ecole Normale Sup´erieure, Paris, France
e Department of Integrative Physiology and Pathobiology, Tufts University School of Medicine, Boston, MA USA
Abstract
Darwin introduced the concept that random variation generates new living forms. In this paper, we elaborate on
Darwin’s notion of random variation to propose that biological variation should be given the status of a fundamental
theoretical principle in biology. We state that biological objects such as organisms are specific objects. Specific objects
are special in that they are qualitatively different from each other. They can undergo unpredictable qualitative changes,
some of which are not defined before they happen. We express the principle of variation in terms of symmetry changes,
where symmetries underlie the theoretical determination of the object. We contrast the biological situation with the
physical situation, where objects are generic (that is, different objects can be assumed to be identical) and evolve in
well-defined state spaces. We derive several implications of the principle of variation, in particular, biological objects
show randomness, historicity and contextuality. We elaborate on the articulation between this principle and the two
other principles proposed in this special issue: the principle of default state and the principle of organization.
Keywords:
Variability, Historicity, Genericity, Biological randomness, Organization, Theory of organisms
Since the beginning of physics, symmetry 2 Invariance and symmetries: physics as the
considerations have provided us with an extremely
domain of generic objects 3
powerful and useful tool in our effort to understand
nature. Gradually they have become the backbone
3 Variation and symmetry changes: biology as
of our theoretical formulation of physical laws.
the domain of specific objects 5
Tsung-Dao Lee 3.1 Randomness proper to specific objects . . . 6
The artificial products do not have any molecular 3.2 Constraints and specific objects . . . . . . . 7
dissymmetry; and I could not indicate the existence 3.3 Constraints and randomness . . . . . . . . . 7
of a more profound separation between the 3.4 Historicity . . . . . . . . . . . . . . . . . . . 9
products born under the influence of life and all the 3.5 Contextuality . . . . . . . . . . . . . . . . . 10
others. 3.6 Variability . . . . . . . . . . . . . . . . . . . 11
L. Pasteur 3.7 Modelization and specific objects . . . . . . 12
3.8 Conclusive remarks on the principle of vari-
ation . . . . . . . . . . . . . . . . . . . . . . 13
Contents 4 Bringing organization into the picture 13
4.1 Organization grounds constraints in specific
1 Introduction 2
objects . . . . . . . . . . . . . . . . . . . . . 13
4.2 The epistemological status of closure under
∗ Corresponding author variation . . . . . . . . . . . . . . . . . . . . 14
Email addresses: mael.montevil@gmail.com (Ma¨ el Mont´evil),
matteo.mossio@univ-paris1.fr (Matteo Mossio),
4.3 Relevant variation with respect to closure . 14
arnaud.pocheville@sydney.edu.au (Arnaud Pocheville), 4.4 Closure remains closure under variation . . 15
longo@ens.fr (Giuseppe Longo)
URL: http://montevil.theobio.org (Ma¨ el Mont´ evil), 5 Non-identical iteration of morphogenetic pro-
http://www.di.ens.fr/users/longo/ (Giuseppe Longo) cesses 16
• Published as: Ma¨ el Mont´evil, Matteo Mossio, Arnaud
Pocheville, Giuseppe Longo, Theoretical principles for biology: Vari-
ation, Progress in Biophysics and Molecular Biology, Available on- 6 Conclusions: back to theoretical principles 17
line 13 August 2016, ISSN 0079-6107, http://dx.doi.org/10.1016/
j.pbiomolbio.2016.08.005.
Preprint submitted to Progress in biophysics and molecular biology August 22, 2016
1. Introduction Such a theoretical framework does not (yet) exist for the
biology of organisms and our proposal aims at contributing
A striking feature of living beings is their ability to to the elaboration of the “biological counterpart” of the
change. All naturalists know that two individuals of the theoretical frameworks and abstract objects at work in
same species usually display important qualitative differ- physics.
ences. All experimentalists know that two replicate ex- It is worth emphasizing that, although we will elab-
periments can give quite unexpectedly different results – orate on the concept of variation by analogy with and
even in the absence of any abnormality in the experimental in contrast to the physico-mathematical perspective, we
setup. by no means advocate a physico-mathematical treatment
Variation took a central role in biological reasoning in of biological phenomena. Rather, we think that biology
Darwin’s book The Origin of the Species (1859) in which in general, and the biology of organisms in particular,
it served as a means to explain the current diversity of requires a significant change of perspective with respect
life, by virtue of the concept of “descent with modifica- to the physical viewpoints and methodologies. Typically,
tion” (Darwin, 1859, pp.119-124): organisms might show physics provide an ahistorical understanding of the phe-
some differences from their parents, these differences might nomena studied2 . In contrast to physics and in line with
be heritable and, under some proper conditions, accumu- the theory of evolution, we argue that historicity is an es-
late to form new lineages. Importantly, to Darwin, some sential feature of biological phenomena and that biological
of these variations would be “chance” variations, that is, historicity stem from the principle of variation.
changes that would be unrelated to the conditions of exis- The principle of variation is related to the other princi-
tence of the organisms, and even unpredictable (Darwin, ples put forward in this special issue: the biological default
1859, p.p 131, 314)1 . In so doing, Darwin introduced con- state (proliferation with variation and motility), and the
tingency and historicity into biological thinking: accidents principle of organization. The default state is described
would happen along life’s trajectory, which would at the as a primary generator of variation; when a cell divides, it
same time be unpredictable, unrepeatable, and have long generates two non-identical cells (Soto et al., 2016). The
lasting effects (Gould, 2002, p. 1334). principle of variation specifies the nature of the difference
In this paper, we elaborate on the Darwinian idea of between these cells. The principle of organization is a way
“chance” variation. We argue that variation should be to interpret biological functions as a property stemming
given the status of a principle in biology, and in particu- from the role that parts play in the maintaining of a sys-
lar organismal biology. Informally, the principle of varia- tem (Mossio et al., 2016, 2009; Mont´evil & Mossio, 2015).
tion states that biological objects (such as organisms) con- According to this principle, a biologically relevant part
tinually undergo modifications. Some of these variations (constraint) both depends on and maintains other parts of
have functional repercussions, which we discuss with pre- the organism, thus forming a mutual dependence (labeled
cise concepts in section 3). Moreover, whatever the math- “closure” for historical reasons). In Mossio et al. (2016),
ematical frame used to describe an object, unpredictable variation and organization are discussed as two intertwined
variations are nevertheless possible: the principle of varia- principles: organization is a condition for variation and fa-
tion thus implies that the existence of exceptions is the rule vors its propagation, whereas variation is a condition for
in biology. However, a proper biological theory cannot be the maintenance and adaptation of biological organization
a mere catalog of exceptions. Accommodating the changes and for the generation of functional innovations. In sec-
biological organisms undergo during their lives (ontogene- tion 4 of this text, we argue that any relevant variation is
sis), as well as during evolution (phylogenesis), in a general a variation of an organization.
theory is a specific challenge raised by biological systems, Biological variation occurs at all levels of organization,
in particular in contrast to physical theorizing. from the molecular level to large scale structures and func-
In physics, theoretical definitions enable us to discuss tions (West-Eberhard, 2003; Dueck et al., 2016). Single
abstractly and adequately the behavior of objects (such cell observations on one side and high throughput tech-
as the trajectory in space of a punctual object of mass m nologies on the other enable biologists to observe both
in classical mechanics, or the behavior of quantum objects inter-cellular and inter-individual variations, which have
as a vector in a Hilbert space in Quantum Mechanics). received an increasing amount of attention (Elowitz et al.,
2002; Collective, 2005; Rivenbark et al., 2013). There
1 This concept of chance variation contrasts sharply with, for in-
stance, the concept of variation of Lamarck (1809) another father of 2 As a matter of fact, physical approaches and methodologies are
theoretical biology. To Lamarck, variations would be directed by the
not confined to the physical and biophysical domain and have, in
conditions of existence. This directedness entails that if the condi-
part, percolated in biology and even social sciences. Such was the
tions of existence re-occur in time, evolution is repeatable and thus,
case, for instance, of the proposal of vital forces by some vitalists.
ahistorical (Gould, 2002, p. 191). Other 19th century writers would
These vital forces were conceived by analogy with Newtonian grav-
advocate that variation would be so canalized (by the properties of
itation and would entail spontaneous generation as a result of this
the organisms) as to direct evolution (when evolution was acknowl-
force acting on the right objects (De Klerk, 1979). Vital forces are an
edged). See e.g. Bowler (2005; Pocheville & Danchin (2016) for more
example of how the physico-mathematical approach typically implies
details.
an ahistorical understanding of the living, as we stress below.
2
are many generators of variation among which are ran- objects are objects which are all of the same kind from the
dom gene expression, instability in morphogenetic pro- point of view of the theory (they typically obey the same
cesses and randomness in biological rhythms. In partic- equations). An apple, the Earth, an anvil, for example,
ular, cellular proliferation generates variation (Soto et al., are all objects with a given mass and center of gravity
2016). As for temporal scales, living systems undergo vari- and, from the point of view of classical mechanics, they all
ation during their lives (ontogenesis), as well as during obey the same equations in the vacuum. Moreover, they
evolution (phylogenesis), and these two aspects cannot be continue to obey the same equation during their dynamics
analyzed independently (Danchin & Pocheville, 2014). In even though they undergo some changes: this is because, in
this paper, we focus on variation as a general feature of physical language, their changes are restricted to changes
biological systems without a privileged level of analysis. of state. Equations are not about specific values of the
This enables us to discuss general features that are proper parameters or states; instead they jointly describe generic
to biology and to stress key differences with respect to relations between parameters, states and the changes of
physics. states4 . This is why changes of state of an object do not
The central implication of this paper is the distinc- affect the validity of the equation which describe its behav-
tion between the objects as conceived in physical theories ior. For example, the mass is an element of the description
(generic objects) and the objects as conceived in biology of some generic objects, formalized by a generic variable
on the basis of biological variation (specific objects)3 . In m representing jointly and synthetically all the possible
what follows, we discuss first shortly what generic objects masses.
are, what kind of manipulation they enable, and how their Physical objects, hence, are generic objects. More gen-
analysis grounds physical theories (section 2). Then, we erally physical ‘laws’ are about generic objects. Consider
contrast generic objects with the variation that biological for example the fundamental principle of dynamics: mass
objects exhibit. We propose that biological objects should times acceleration equals the sum of external forces applies
be understood as specific (in section 3). Specific objects to the object. Here, the “external forces” are understood
are, in particular, fundamentally historical, variable and in a completely generic manner and any kind of forces may
contextual. Thus, the specificity of organisms encompasses be involved.
biological individuation and diversity. We also discuss the Typically, a physical object is described in a mathe-
interplay between specific objects and physical morpho- matical space which is generated by the various quantities
genesis. Then, in section 4 we elaborate on the integration required to describe this object. This mathematical space
between the principle of variation and the principle of or- is called the ‘phase space’5 . In classical mechanics, the
ganization, between the notion of biological specificity for phase space is the space of positions and momenta. This
biological objects and the notion of organization and “con- mathematical space is given in advance; it pre-exists the
tingent genericity” (Moreno & Mossio, 2015; Mont´evil & description of the object. The behavior of the physical
Mossio, 2015; Mossio et al., 2016). Finally, we develop the object is defined as the way in which the object changes
idea that biological systems are characterized by the non- in its phase space. The space is also assumed to provide
identical iterations of morphogenetic processes (section 5). all the causes of the changes of the object, and thus it
specifies the quantities that should be measured experi-
mentally. In classical mechanics, positions and momenta,
2. Invariance and symmetries: physics as the do-
in combination with properties such as the mass, are the
main of generic objects
quantities required to understand the changes of positions
The principle of variation poses novel challenges with and momenta over time.
respect to how mathematics enables us to describe the A phase space, however, is not sufficient to understand
world. To better identify these challenges, we first make a the behavior of an object because the quantities it pro-
detour by physics and show the role mathematics play in vides need to be articulated together to understand the
physical theories. changes of the object6 . In physics, a theoretical framework
Physics is based on mathematized theories. Histori- requires equations that depend on the variables symboliz-
cally, the development of physical theories has been inter-
twined with the development of appropriate mathematics 4 Simulations suffer from a shortcoming in this respect. While a
to frame and define their objects: they have “co-evolved”. program does describe generic relationships between the variables, a
We submit that mathematized physical theories rely on simulation run only provides one trajectory for specific values of its
input. Whether this trajectory is representative of the behavior of
the manipulation of generic objects (Bailly & Longo, 2011; the system for other values of the input, that is to say whether the
Longo & Mont´evil, 2014a). The notion of generic objects is behavior obtained is generic or not, is a very difficult mathematical
abstract, as it lies at the core of physicomathematical rea- issue (Stoer & Bulirsch, 2013).
5 Some physicists restrict the notion of ‘phase space’ to positions
soning. However, the intuitive idea is quite simple: generic
and momenta. Here, phase space means in general the space of
mathematical description of the object.
6 The a priori diversity of possible trajectories in such a space is
3 An introduction to this distinction is given by Soto & Longo
unfathomable in the sense that no axiomatic is sufficient to describe
(2016). all their possible mathematical features.
3
ing the quantities describing the objects. The behavior of Phase space
an object, that is to say its changes, is determined as a ho
sts
fy
specific trajectory by equations that single it out in the
ri
e
at
ve
phase space. Equations are valid for the phase space (or,
ner
ge
at least, some regions of it) and depend on its quantities. trajectories
The behavior of the object is completely determined by Symmetries
ve
the quantities that define its phase space and the corre- rif e
y m in
ju
sponding equations. Predicting a trajectory corresponds sti ter
to making this trajectory mathematically remarkable. To fy de
Equations
this end, equations typically correspond to optimization
principles (for energy, entropy, entropy production, etc.),
which enable physicists to single out a trajectory, the op- Figure 1: Articulation of the different components defining a generic
timal trajectory that the system follows according to the object in physics. Equations determine the trajectory of a system,
theory. Optimization principles and the ability to derive and this trajectory takes place in a mathematical space. Both the
equations and the space have a structure that is described by the the-
equations are essential for fundamental physical theories oretical symmetries that frame the object and that are valid by hy-
and special models to make predictions. pothesis. There is a fundamental feedback that we do not represent
For the purposes of this paper, the key question to be here: trajectories are the endpoint which fundamentally justifies the
asked at this point is what justifies the use of the spaces whole theoretical construction of the generic object as experimenter
can observe them.
and equations in the theoretical constructions of physics.
In part, these mathematical structures stem from axioms
and are justified by their consequences. However, there aspects of an object invariant. For example, cutting an
is more to say on the nature of fundamental hypotheses apple into two halves does not change the way it falls in
of physics and the way in which they justify the use of the vacuum. Hence, an apple and its two halves are sym-
mathematics. metric (they are the same) from the point of view of free
Because whole classes of concrete objects are described fall in classical mechanics. Allometric relationships pro-
in the same mathematical frame, they are studied as the vide a biological example of symmetry (Longo & Mont´evil,
same generic object, and all have the same behavior. As 2014b). In mammals, the average period of rhythms such
we evoked above, a piece of lead, an apple, or a planet are as heart rate or respiratory rate is found experimentally
all the same objects from the viewpoint of classical grav- to depend on mass with the relation τ ∝ M 1/4 . Measur-
itation: they all are point-wise objects with a position, a ing such relationships amounts to assuming that the basic
momentum, a mass, and they all are subject to the prin- properties of metabolism are preserved under the trans-
ciple of inertia and gravitational forces, described by the formations consisting of changing sizes and species, and
same equations. In this respect, there is no relevant dif- thus that mammals of different masses are symmetric as
ference between them and they are described jointly and for their internal rhythms (West & Brown, 2005; Longo
synthetically as the same generic object. At the core of & Mont´evil, 2014b). Lastly, the assumption that different
this approach to natural phenomena lies the identifica- replicates of an experiment enable us to access the same
tion of non-identical objects. This identification of non- situation also corresponds to an assumption of symmetry
identical objects is made explicit by transformations that between the replicates: they are all supposed to behave in
leave these objects invariants (i.e. symetries). Putting fundamentally the same way8 .
an emphasis on transformations is a modern approach in Symmetries are the basis of the mathematical struc-
mathematics and physics that we build upon in this paper. tures in physics; that is the phase space and the relevant
In particular, invariants are best described by the trans-
formations that preserve them and which make explicit a
biology. For instance, it is possible to deform a balloon into a sphere
mathematical structure. or a rod shape without tearing and/or stitching, but it is impossible
Generic objects are, for the most part, defined by the to transform it into a donut. Similarly, it is impossible to transform
transformations that preserve them, and that enable us to a cell into two cells without tearing and/or stitching the membrane,
define stable mathematical structures. We call such trans- where stitching corresponds here to the fusion of the membrane by
pinching, and tearing corresponds to the final separation of the cells.
formations ‘symmetries’. The notion of symmetry we use In all these cases, continuous deformations are considered as symme-
is more general than the concept of geometrical symme- tries, insofar as they preserve topological invariants and, reciprocally,
try in a three-dimensional space. Yet, the underlying idea the topological invariants are the ones preserved by continuous de-
formations. As a result, one can define different categories of shapes
is the same: geometrical symmetries are transformations
on the basis of their inter-transformability. Continuous deformations
which leave a geometrical figure invariant. Rotating a cir- fall under our concept of symmetry and are characteristic to the field
cle around its center, for instance, does not modify the cir- of topology.
8 Notice that such an assumption is required in order to perform
cle: it verifies a central symmetry7 . Similarly, symmetries
statistical analyses. The most common statistical assumption is that
(in general) are transformations which leave the relevant two variables are identically distributed, that is to say that the two
considered situations are symmetric as far as their probability dis-
7 Another example comes from topology, a notion very useful in tributions are concerned.
4
equations. Accordingly, they constitute fundamental phys- view, qualitative changes typically imply changes of the
ical assumptions which are less anthropomorphic than the relevant mathematical structures and, accordingly, changes
notion of law and more meaningful than conservation prin- of symmetries. For example, changes of states of matter in
ciples (see for example Van Fraassen, 1989; Bailly & Longo, phase transitions typically correspond to changes of sym-
2011; Longo & Mont´evil, 2014a). metries: a liquid is symmetric by rotation while a crystal
For instance, the choice of an origin, three axes and a is not, because of its microscopic structure (see figure 3).
metric are mandatory in order to write equations and per- In the biological domain, the organization of any cur-
form measurement of positions and velocities (in Galilean, rent organism has been shaped by permanent qualitative
special, or general relativity). Although different choices changes, that is, through changes of symmetries. A given
are possible, the consistency of the theory depends on the biological organization is determined by an accumulation
fact that the trajectories obtained in different reference of changes of symmetries both on the evolutionary and the
systems are, in a fundamental sense, the same: in par- ontogenetic times13 . These changes correspond to changes
ticular, they are invariant under suitable classical or rel- in the manner in which functions are performed, or even
ativistic transformations of the reference system. Thus, to the appearance or loss of functions.
the equations of physics are symmetric under these trans- Acknowledging that organisms can vary in this strong,
formations9 . In general, the same trajectory should be functional sense, is not trivial: historically, the preforma-
obtained before and after transformations which are fun- tionists (as for development), and the fixists (as for evolu-
damental symmetries in the theory 10 , and these symme- tion) have held just the opposite view. If the homunculus
tries enable us at the same time to formulate and justify is already in the egg, or, in modern terms, if dna already
the equations and the phase space 11 . contains a blueprint of the organism, then development is
In short, physical objects are understood as generic just the unfolding of an already existing organism (with all
objects that follow specific trajectories. Theoretical sym- its relevant properties and functions). Similarly, if species
metries ground this approach to natural phenomena. The do not change over geological time, then obviously organ-
epistemological structure of generic objects is summarized isms conserve the same functions.
in figure 1. In the next section, we discuss the princi- The idea that biological objects genuinely develop and
ple of variation and the major challenges that biological evolve over time corresponds to the idea that the mathe-
variation raises when one tries to frame biological objects matical structures required to describe them also change
theoretically. over time. Thus, stating that development and evolution
involve symmetry changes constitutes nothing more than
a mathematical interpretation of the departure from the
3. Variation and symmetry changes: biology as the
preformationist or fixist stances of development and evo-
domain of specific objects
lution. Evolution is rarely considered as entirely deter-
A central and pervasive property of biological systems mined as the unfolding of historical necessities. Similarly,
is their ability to change their organization over time12 . development should not be seen as the unfolding of a pre-
These changes are not just quantitative changes, they are constituted organization but instead as a cascade of folding
also qualitative. From a physico-mathematical point of leading to the setting up of an organization (figure 2 and
4).
The crucial consequence of this view is that, because
9 Similarly, in electromagnetism the choice of assigning negative
of their permanent symmetry changes, biological objects
or positive charges to electrons is arbitrary; therefore, permuting the
sign of charges has to leave the equations invariants (the derived
should not be considered as generic objects. Organisms
trajectories remain the same). are not well defined as invariant under transformations.
10 In a mathematical model, some symmetries are theoretical sym-
When an organism is transformed, and in particular when
metries which cannot be violated while others are more pragmatic the flow of time operates on it, the organism may undergo
symmetries that correspond to a particular situation. The two things
should not be conflated. For example, a theoretical symmetry is the unpredictable qualitative changes. As a result, biologi-
assumption that all directions of the empty space are equivalent. cal objects are not well described by the virtuous cycle
However, in a particular setting, all directions may not be equiva- described in figure 1. Accordingly, trajectories are not en-
lent, for example because of the position and the gravitational field tirely framed by a mathematical framework: they may es-
of some planets. Another theoretical symmetry is the symmetry be-
tween positive and negative charges in classical electromagnetism. cape such frameworks and require a change in the symme-
11 Such justification of equations by symmetries is, in particular, tries, space of description, and equations used to describe
the core of Noether’s theorem, which justifies the conservation of the object (figure 2).
energy (resp. momenta) on the basis of a symmetry by time (resp.
We propose then to understand biological objects (and
space) translation of fundamental equations, among many other con-
served quantities (Byers, 1999; Longo & Mont´ evil, 2014c). organisms in particular) as specific objects 14 . Specific ob-
12 While we mean here ‘organization’ in the technical sense dis- jects are constituted by a particular history of relevant and
cussed in Mossio et al. (2016), the reader can also interpret the
notion in a more informal manner. The different parts of an or-
ganism depends on each other and form a coherent whole. This 13 A more detailed presentation of most of these ideas can be found
interdependence of the parts and their relation to the whole form in Longo & Mont´ evil (2014a) and Longo & Mont´ evil (2011, 2013).
the organization of organisms. 14 Our concept of specificity should not be confused with other
5
Phase space Phase space change
ve
es
rifi es
ge
re
fy
ca
ne
ui
ri
te
es
pe
ve
ra
ra
req
t
ne
trajectories Symmetry
ge
Symmetries change
ve es
re
rif fi
qu
pe
y r i
ire
ve s
ca
jus
fie
es
tif t i
y Equations Equation change us
j
Figure 2: Scheme of an elementary symmetry change in biology. An initial situation, on the left, is described by analogy with physics (see
figure 1). However in biology, variation can escape such a frame. Understanding the object then requires a change of symmetry and of the
whole mathematical structure framing the object. Trajectories are at the center of this change, they escape the initial frame and thus require
a change of the symmetries describing the object.
unpredictable symmetry changes over time, at all time- be defined generally as unpredictability with respect to a
scales. Specific objects can be understood as the opposite theory. The notion of randomness which stems from the
of generic objects: two instantiations of a specific object principle of variation is not endowed with a probability
may always differ by at least one of their relevant qual- measure.
itative aspects (in a given theoretical frame), while two Let us first characterize randomness in the case of a ba-
instantiations of a generic object do not. For example, sic symmetry breaking, typically encountered in physical
two organisms, be they clones, may always differ in one of models. Let us start with a situation which is symmetric,
their relevant qualitative properties, for instance because for example a gas (figure 3, top). All directions are equiv-
they may have undergone differences in their morphogen- alent for this object: all macroscopic quantities (density
esis, i.e., they have been constituted by different develop- of the gas, pressure, etc.) stay the same after rotation.
mental histories. When the symmetry is broken, directions are no longer
On the basis of the concept of specific objects, we can equivalent; for example, there are privileged directions cor-
now state the principle of variation: responding to a crystal structure after a phase transition
(figure 3, bottom). The symmetry of the initial situation
Principle of variation: means that all directions are initially equivalent and then
Biological organisms are specific objects. that it is not possible to deduce the subsequent privileged
The principle implies that biological organisms undergo directions in the crystal. As a result, the directions of the
changes of symmetry over time and that, as we discuss be- crystal are random in this theoretical account. Moreover,
low, some of these changes cannot be stated in advance15 . since all directions are symmetric in the initial conditions,
In other words, the mathematical structure required to all directions have the same probability to become one of
describe organisms is not stable with respect to the flow the crystal’s privileged directions.
of time. Qualitative changes of structures and functions This physical situation exemplifies how symmetry break-
occur over time and some of them are unpredictable. ing and randomness are associated and how the initial
We now expand on several aspects and implications of symmetries define and justify probabilities (see Longo &
the principle of variation. Mont´evil, to appear, for a general analysis of this associa-
tion).
3.1. Randomness proper to specific objects Symmetry breaking and the associated randomness are
A fundamental feature of the principle of variation is relevant for biology but we submit that they are not suf-
that it includes an original notion of randomness: the very ficient. Biological randomness includes a fundamentally
fact that biological objects undergo unpredictable symme- different notion. In the above case, the possible outcomes
try changes. Generally speaking, the notion of randomness (all the possible directions in three dimensions) are defined
is often conflated with the idea that events have some prob- before the symmetry breaking, as it is the mathematical
ability of occurrence. However, scientific approaches to space on which symmetries act. Saying that the gas is
randomness are richer than the notion of (classical) prob- symmetric by rotation requires us to define rotations and
abilities (see for example Longo et al., 2011, for a discus- therefore the set of all possible directions on which rota-
sion at the crossroads of different fields). Randomness may tions act. In biology, in contrast, the principle of variation
poses that the list of possible outcomes and therefore the
concepts of ‘biological’ specificity, such as chemical specificity of en- relevant symmetry changes are not pre-defined. For exam-
zymes, or causal and informational specificity (see Griffiths et al., ple, it is not possible to embed all the spaces of description
2015). of current and future organisms within the space of de-
15 We would argue that even the rate of possible symmetry changes
scription of the last universal common ancestor (LUCA).
cannot be stated in advance.
6
3.2. Constraints and specific objects
= The principle of variation does not preclude the pres-
ence of elements of stability in biological systems. On the
contrary, in order to show experimentally and describe the-
oretically a change of symmetry, the preceding and follow-
ing situations have to be stable enough to be described. In
6= other words, a set of symmetries has to be at least approx-
imately valid long enough before it changes for an observer
to discuss it and after the change the new set has to be
met for some time too. For example a given geometry
Figure 3: Example of a symmetry breaking. The left pictures corre- of bones is conserved during movements of the organism
spond to an initial situation and the right ones to the same situations at short time-scales, which corresponds to the conserved
after a rotation (represented by the arrows). The above diagrams
show a disordered situation such as a gas or a liquid. This situa- symmetry of a solid (the relative positions of points in a
tion is statistically symmetric by rotation, there are no privileged solid do not change). However, this geometry is plastic
directions. By contrast, the situation below corresponds to a crys- at longer time scales and very important changes can oc-
tal such as graphite. It is not symmetric by rotation (except with
cur especially during development (West-Eberhard, 2003).
an angle of 180◦ ) and it thus has directions which have an intrinsic
physical meaning. The transition from the situation above to the The change of two bones geometry at different times thus
situation below implies the introduction of new relevant elements: corresponds to a symmetry change, but the symmetries of
the directions of the crystals, which are random. these bones are met at short time-scales.
We call constraints the relevant stable elements at work
in biological systems and their associated symmetries. Con-
A part of the relevant symmetry and symmetry changes
straints are local stable elements, in the sense that they
can only be listed a posteriori, that is, after their real-
only concern a particular aspect of a given organism. In
ization. These changes only make sense as a result of a
addition, constraints are contingent insofar as they, and
previous history. Not only lineages, but also individual
their associated symmetries, may change over biological
organisms, are subject to biological randomness, as their
time (which is implied by the principle of variation).
development can sometimes take new routes which were
In short, we define constraints as symmetries (i.e. sta-
not expected in advance (e.g. West-Eberhard, 2003).
ble mathematical structure) witch have a restricted range
Note that we consider symmetry changes in general
of validity and are used to describe a part of a specific
and not just symmetry breaking. Symmetry breaking cor-
object.
responds to symmetry changes which start from a situation
that respects a given symmetry to a situation where this
3.3. Constraints and randomness
symmetry is no longer valid, as discussed above. Other
symmetry changes are possible, for example one can go In this section, we discuss the articulation between two
from an asymmetric situation to a symmetric one. In bi- kinds of randomness in specific objects. This discussion is
ology, symmetry changes include the appearance of new more technical and may be skipped in a first reading.
and unpredictable symmetries corresponding to new rele- A constraint (or a combination of constraints) exerted
vant parts and their functioning. For example, the appear- on biological dynamics may lead to a situation in which
ance of sexual reproduction in evolution corresponds to a symmetry changes (if any) occur in a generic manner,
separation of individuals in two genders in many species, typically as symmetry breaking. In the case of generic
where new symmetries (or equivalence) between males on symmetry changes, these ‘random’ changes can be stated
one side and females on the other become fundamental in advance, even though their specific outcome cannot.
as for their role in reproduction. New associated variables This randomness can be derived from constraints, and it
become relevant, for example the sex ratio of a population. is weaker than the randomness proper to specific objects.
Because of symmetry changes, the phase spaces of bi- Let us start with morphogenesis as an example. Most
ological objects also change in unpredictable ways over (if not all) mathematical models of morphogenesis involve
time. Symmetry and phase space changes constitute a a symmetry change, which usually is a symmetry break-
specific form of randomness, proper to biological systems ing. Consider for instance Turing’s model of morphogene-
(Longo & Mont´evil, 2012; Longo et al., 2012a; Kauffman, sis (Turing, 1952)16 . In this model, the equations describ-
2013; Longo & Mont´evil, 2013). Biological randomness ing reactions and diffusion of chemicals remain invariant,
typically manifests itself through the appearance of new so that their properties (rate of reactions, coefficient of
relevant quantities, parts, functions, and behaviors over diffusion, etc.) are stable constraints. In turn, these con-
time (for example limbs, toes, toenails, all the quantities straints lead together to a symmetry breaking, because of
required to describe them and the various functions that
they can have). 16 Turing’s model is based on a basic symmetry breaking, where a
situation that is initially symmetric by rotation forms a pattern of
alternation of concentrations of chemicals (and new quantities are
needed to describe where this pattern is located).
7
the sensitivity of the non-linear dynamics to initial con- given space of possibilities which may be given a priori
ditions (an instability, says Turing): minor fluctuations probabilities. Biological objects are — by hypothesis —
trigger different outcomes. specific, but when we describe a particular change of sym-
Another very different example of biological symmetry metry, it is studied a posteriori as a generic aspect of the
breaking is the dna recombinations in the maturation of object, and can be added to the past possibilities of a sys-
lymphocytes (Thomas-Vaslin et al., 2013). The random tem. Randomness is then not correctly framed by a priori
process of recombination in a cell can be seen as a symme- probabilities. Probabilities, if any, are defined a posteri-
try breaking from a situation where all the recombinations ori. A specific possibility is accommodated by the space
to come are equivalently possible to a situation where only of possibilities, but this space is obtained a posteriori and
one recombination is actually realized in each cell. After obviously does not include all future possibilities.
recombinations, the description of the system has to in- Let us unpack this idea. A physical symmetry breaking
clude which possibility each cell has “chosen”. This sym- is a simple elementary process: a symmetry is met by the
metry breaking makes the diversification of the immune system, and after the symmetry breaking event, the sym-
repertoire possible under the constraint of enzymes. metry is no longer met. The possible breakings are given
Both cases (morphogenesis and dna recombinations) by the initial set of symmetries and make mathematical
involve stable constraints, in an extremely sensitive pro- sense when they can be described in a given mathematical
cess, which leads to a change of symmetry. These con- space where the symmetry operates. However, if a situ-
straints are stable parts of the organization of the consid- ation is and always has been completely symmetric, the
ered organisms. As a result, the associated changes are symmetries do not change anything and thus, cannot be
robust in the sense that they will occur as a consequence properly evidenced as transformations (because the object
of these constraints. In such situations, a generic change of is not changed at all). Thus the logic required to describe
symmetry is established, which generates “new” relevant a new symmetry breaking has two steps. First the sym-
quantities but in a generic manner, i.e. the change belongs metry that will be broken has to be added to the initial
to a set of predefined possibilities. These new quantities definition of the system and accordingly the states that are
are new in a weaker sense than the unpredictable new di- initially symmetric have to be added to the phase space
mensions of description that specific objects can generate. of the object. They are added because they are required
For example, the recombinations in the immune system to accommodate their future breaking. Then, and only
can be seen as generic, as a set of possible physico-chemical then, may the symmetry be broken. Such a modeling is
recombinations of molecules. The outcome of such recom- retrodictive: the mathematical space, needed for an equa-
binations is probably unique because the odds of perform- tional model, can be given only after the change has been
ing the same recombinations twice are vanishingly small, observed. In general, then, a biological dynamic must be
but this outcome is still generic. The situation is anal- understood as a possible path, out of many established
ogous to the physical case of the positions of individual along the biological dynamics, which consists in the com-
molecules in a gas which are basically unique, whereas the position of stepwise symmetry changes.
gas is still in a generic configuration because the gas is in In a given situation, some symmetry changes can be
a configuration of maximum entropy. However the actual spelled out and analyzed in a generic framework because
immune repertoire in an adult mammal is not fully deter- they are stabilized by (local) constraints. Let us consider
mined by the generic properties of recombinations because such an elementary biological symmetry change, for exam-
the recombinations are just a part of the process estab- ple in a morphogenesis model. We can describe it explicitly
lishing this repertoire. The immune repertoire strongly with generic constraints but it is also possible to leave it
depends on the specific history of the given organism, its implicit and consider that this single symmetry change is
environment, non-genetic inheritance (through milk and taken into account by the specificity of the object, among
the microbiome), etc. (Thomas-Vaslin et al., 2013). The many other changes. The choice depends on the perspec-
immune repertoire has a causal structure that is not deter- tive adopted to understand a given situation, including the
mined by pre-existing regularities. The dependency on the scale of description and the phenomena of interest. For ex-
organism’s history is functional, it determines the immune ample, the intestine folding are usually kept implicit when
response to specific pathogens and contributes to the dy- studying brain morphogenesis.
namic relationship with the microbiome. The biologically Even though the boundaries of specific and generic as-
relevant properties of the immune repertoire are not the pects of an organism are relative and may change after a
generic properties of recombinations, instead they are the new possibility is acknowledged or as a result of a change
specific properties which stem from a history. Hence, the of perspective, the accurate description of any biological
actual repertoire of the adult contains more meaningful organism will always involve a component of specificity. In
novel structures than the initial probabilistic recombina- a given representation of an organism, all changes of sym-
tions. metry are then either accommodated by the specificity of
Now, every time we describe a symmetry change ac- the object or by generic symmetry changes. The concept
cording to current physico-mathematical methodology, it of the specificity of biological objects aims to enable us
takes a generic form, that is, a possible change in a pre- to take into account theoretically all symmetry changes
8
without spelling out all of them explicitly. their time is that of a process. Their historicity is embed-
ded within a pre-defined phase space.
3.4. Historicity The fact that we can understand such spontaneous ob-
Historical objects are objects whose properties are ac- jects on the basis of a stable generic mathematical struc-
quired or lost over time, and cannot all be described ahead ture is not fortuitous. Indeed, their spontaneous charac-
of time. The fact that biological organisms are specific ter corresponds to the fact that these objects can emerge
objects straightforwardly implies that they are historical from homogeneous initial conditions in the mathematical
objects and, in particular, contingent objects in Gould’s framework used to describe them. By contrast, specific
sense (Beatty, 1995; Gould, 1989). Historicity thus goes objects are not framed by stable mathematical structures:
hand in hand with biological randomness, which corre- they cannot be derived from homogeneous initial condi-
sponds to the fact that a situation after a random event tions and cannot be obtained spontaneously in practice.
cannot be stated with certainty before the event. Thus, Even in the “origin of life” field, the aim is to produce
a system showing biological randomness shows historicity: a cell which can evolve and not a cell that is similar to
the object takes a particular path among several possible all current cells as they have evolved for billions of years.
paths through time. Reciprocally, historical objects nec- Moreover the aim is certainly not to obtain a cell similar
essarily show some randomness. to any specific species (Pross & Pascal, 2013).
Let us first consider an analogy with dynamic systems. According to the principle of variation, biological ob-
We can see a trajectory defined by a differential equation jects are the result of a cascade of unpredictable symmetry
as the sum of infinitesimal changes from the initial condi- changes, which implies that they do not follow optimiza-
tions to any time point. By analogy, it is conceivable to see tion principles and that they are not spontaneous. To be
biological historicity as a sum or a sequence of variations sure, biological objects did appear spontaneously in the
since the origin of life. However, this idea does not have history of life, but should one re-run the history of the
a well-defined mathematical and theoretical sense, insofar Earth, one could not expect to obtain the same biological
as such a history is not entirely accessible. Nevertheless, objects. It is not even possible to state in advance the
it is still possible to clarify the present in the light of the mathematical space of possible forms that could be ob-
past — and, as a matter of fact, this is precisley one of the tained. The historicity of biological objects is not embed-
aims of evolutionary theories. ded within the phase space anymore (as it was in physics):
As discussed in Longo et al. (2015), although histor- rather, the principle of variation means that the phase
ical objects exist also in physics, they are historical in a space itself is historical (figure 4).
weaker sense. Self-organized physical objects, for instance, At first sight, though, the claim that the phase spaces
are sometimes described as historical, mostly because they in biology are historical seems too strong: aren’t there
depend on a symmetry breaking. For example, the appear- some aspects of biological objects which are ahistorical?
ance of convection cells in a fluid corresponds to a qualita- Evolutionary convergences, for instance, seem to be an ex-
tive change in the macroscopic dynamics of the fluid. Nev- ample of an ahistorical aspect of the living: convergent
ertheless, self-organized objects are spontaneous: they can features seem to be obtained independently of (some as-
be obtained de novo. Theoretically, they can be described pects of) the past history of the organism. Let us first
as the spontaneous self-organization of flows of energy and point out that evolutionary convergences are not about
matter. Even the physical situation of the early history of invariant properties of a given object over time, they are
the universe can be obtained experimentally “just” by tun- about mathematical structures that are similar in differ-
ing a parameter (by obtaining very high local densities of ent historical paths. Let us consider the case of the cam-
energy with particles accelerators)17 . era eye of the vertebrates and of the cephalopods as an
Despite these analogies though, physical self-organizing example. These eyes have different evolutionary origins
processes have no historical or evolutionary time in a strong but they are nevertheless similar and one could argue that
theoretical sense; they may just have the time of a process. they would be instances of the same generic object from a
They entirely obey optimality principles from physics and physicomathematical viewpoint, when described in terms
past events have not shaped their properties, insofar as the of optical geometry for example.
symmetry breakings that self-organizing processes may en- The principle of variation, however, implies that the
counter are all pre-defined within the theory. A hurricane convergence is very unlikely to be qualitatively exact. There
does have, so to speak, a “birth”, a “life”, and it does even- would always be a relevant biological description which
tually “die out”; yet, hurricanes have been the “same” kind would distinguish them sharply by pointing to differences
of object for the past four billion years on Earth. Again, in their organization and in their articulation with the rest
of the organism. For instance, the retina is inverted in ver-
tebrates: the axons of photoreceptors and their connection
17 Incidentally, the idea of spontaneous generation in biology
to ganglion cells and the optic nerve are located between
stemmed from the same kind of reasoning: (generic) biological ob- the receptors and the light source, creating a blind spot at
jects would appear spontaneously by self-organization in the appro-
priate milieu (De Klerk, 1979).
the level of the optic nerve. In cephalopods, axons are be-
hind the photoreceptor which does not create such a blind
9
spot. A close analysis of both the phylogenetic and the on- Let us discuss two examples of internalization of the
togenetic paths makes the difference understandable: the context on the developmental and on the evolutionary time
high modularity of the cephalopods’ brain derives from an scales, to show how it can lead to unexpected behaviors of
early separation of the brain’s modules by an invagination biological objects.
of the ectoderm, in contrast to the evagination of the dien- On the developmental time-scale, an example of inter-
cephalon, due to the late separation of the eye component nalization of past contexts is provided by the response of
of vertebrates’ brains. cells to hormones (Soto & Sonnenschein, 2005). Basically,
In short, the principle of variation implies that strict the response of a cell to hormones does not depend only
evolutionary (or developmental) convergence never occurs: on the specific receptor and corresponding hormone in-
symmetry changes are such that biological objects drift in volved but, rather, on the developmental history of this
a burgeoning phase space, and partial convergences always cell. More precisely, precursor erythroid cells are expected
embed hidden differences which may be of importance with to differentiate into red blood cells when their erythropoi-
regard to the considered behavior of the biological object etin receptors bind with erythropoietin. However, precur-
in that phase space. Reciprocally, the similarity between sor erythroid cells which have been engineered to lack ery-
the organizations of different organisms stems from com- thropoietin receptors and instead have receptors for pro-
mon descent, that is to say from a shared history. lactin do differentiate into red blood cells when they are
exposed to prolactin, a hormone associated with lactation
3.5. Contextuality (Socolovsky et al., 1997). Conversely, mammary epithelial
Organisms are contextual objects. In our theoretical cells can be engineered to have a hybrid receptor with an
framework, the symmetries of organisms depend on its en- extracellular part of a prolactin receptor and an intracel-
vironment — both on its immediate environment and the lular part of an erythropoietin receptor. These engineered
environments encountered in its past history. cells respond like normal mammary epithelial cells to pro-
The fact that the symmetries of an organism depend on lactin (Brisken et al., 2002). These examples show that it
its immediate environment constitutes another similarity is not the molecular specificity of a signal binding to a re-
with self-organizing physical systems mentioned above, as ceptor that determines the response of a cell to a hormone.
the latter strongly depend on their boundary conditions. In contact with a hormone for which it has a receptor, a
However, the principle of variation makes the contextuality cell rather responds according to the context of its cellular
of biological objects more fundamental than that of phys- lineage during development, that is its trajectory in time
ical systems. Contrary to physics, the possible changes of and space (Soto & Sonnenschein, 2005).
symmetry due to a change of the context are not all pre- On the evolutionary time-scale, a component of an or-
defined. This means that an organism in a new environ- ganism, as a result of a history, may be used for differ-
ment may undergo unpredictable reorganizations, which ent purposes in different contexts. The phenomenon of a
correspond to different relations between its internal con- character (be it the result of past natural selection or not)
straints and the environment, as well as different relations which is coopted for a current use has been named ‘ex-
between its internal constraints, tout court. For example, aptation’ by Gould & Vrba (1982). They provide many
we do not know a priori the many changes that can occur key examples, for instance: “the jaw arises from the first
when bacteria that used to live with many other species gill arch, while an element of the second arch becomes,
in their natural and historical environment are grown as in jawed fishes, the hyomandibula (suspending the upper
an isolated strain in laboratory conditions. Similarly, it is jaw to the braincase) and later, in tetrapods, the stapes,
always difficult to assess whether the behavior of cells cul- or hearing bone” (Gould, 2002, p.1108). An ex-aptation
tured in vitro is an artifact of in vitro culture, or whether is a re-interpretation, or re-use, of a trace of the past in
it is biologically relevant (meaning that it corresponds to a new context and, therefore, cannot be derived from the
a behavior that happens in the context of the multicellular initial function of the parts involved. As a consequence,
organism from which they were taken, see Mont´evil et al., the detailed structure of the internal ear can be better
2016). understood by looking into the cumulative history of ex-
The contextuality of biological objects is coupled with aptations.
their historicity (Miquel & Hwang, 2016): biological or- In light of the principle of variation, the internalization
ganizations tend to maintain the effects of former envi- of current and past contexts provides one way (although
ronments and may even internalize their relationship with not the only one) in which symmetry changes can occur
the environment over time. This holds at the develop- throughout the history of an organism. As an illustration,
mental scale (think of how early plastic responses to the the internalization of the context contributes to explaining
environment might be ’frozen’ later in development, see the difficulty of replicating biological experiments, insofar
also Gilbert & Epel (2009)), at the scale of several gen- as aspects of an experimental situation which can be rel-
erations (for example through epigenetics), and at longer evant to the studied behavior may not be measured and
evolutionary scales (think, for instance, of the presence of can be traces of an (unknown) past (Begley & Ellis, 2012).
lungs and lack of gills in marine mammals, which reflects
a past terrestrial life).
10
Symmetries change Symmetries Besides the flow of time, the second set of transforma-
(Constraints) (Constraints) tions relevant to variability are the permutations of dif-
are a part of
are a part of
ferent organisms or different populations. Permutations
establishes
establishes
correspond to the interchanging of different objects. They
are fundamental symmetries in many physical frames: for
determines time example, it is axiomatic that all electrons follow the same
History Specific object Specific object
randomness equations (but they can be in many different states). In
s
s
ne
ne
experimental biology, permutations of different animals or
mi
i
rm
ter
cells are often assumed to be symmetries: when one con-
te
de
de
siders different animals of a control group, a common as-
Context Context
sumption is that they behave in the same way and that
the quantitative variation observed stems from a proba-
Figure 4: Biological objects and their theoretical structure. Spe- bility distribution that would apply to all of them. This
cific objects are not defined by invariants and invariant preserving assumption, in one form or another, is required to apply
transformations. Instead, specific objects such as organisms undergo
random variation over biological time. Their behaviors are not given
theorems of statistical analysis.
by a synchronic description. Instead, they depend on a history and a According to the principle of variation, however, the
context. Constraints are restricted invariants and symmetries, which permutation between these organisms cannot be taken as
may change over time and frame a part of the behavior of specific a symmetry. Of course, organisms are related by a shared
objects. Experiments and mathematical models usually investigate
constraints and their changes.
history, which enables us to determine that they are mice,
rats, etc., of a given strain. Yet, the transformation which
replaces one organism by another in the same group cor-
3.6. Variability responds to a comparison between the results of divergent
The principle of variation underlies biological variabil- paths stemming from a shared history. Here, divergence is
ity: the fact that multiple organisms or the same organism taken in a strong sense and implies symmetry changes and
or lineage at different times exhibit differences when com- not mere quantitative changes conserving the same sym-
pared to each other. metries. For example, qualitative behaviors differ between
The flow of time is the most fundamental transforma- different strains of the same species, even in unicellular
tion acting on biological objects: as we argued, biological oganisms (Vogel et al., 2015). Now, we illustrate this idea
symmetries and accordingly biological organizations are with a historical example.
not preserved as time passes. At the end of the 19th century, Sir Francis Galton, one
Variability tends to be stronger when considering large of the founders of the notion of heredity, came up with a
evolutionary time scales than for shorter time scales. When device, known as the bean box or the quincunx (see figure
one follows the succeeding generations from the LUCA to 5). The quincunx facilitated the simulation of a binomial
a randomly chosen current organism, for example a rat, distribution (the device would be used to simulate “normal
many relevant aspects of the description needed to under- variability”, Galton (1894, pp.63f)). The device consisted
stand these organisms appear and disappear through time. in a vertical frame with three parts: a funnel in its upper
Variability is also significant at physiological time scales, part, rows of horizontal pins stuck squarely in its middle
even at those that are much shorter than the lifespan of part, and a series of vertical compartments in its lower
the considered organism. Heart rate, for example, does not part. A charge of small items (say, beans or balls) would
obey homeostasis stricto sensu: the beat to beat interval is be thrown through the funnel, travel through the pins,
not invariant (in a healthy situation), and it does not even possibly bouncing in any direction, and would be gathered
display fluctuations around a stable average value. In- by the vertical compartments at the bottom (where they
stead, the beat to beat interval fluctuates in a multiscale would not move anymore). In the end, the distribution of
manner (West, 2006; Longo & Mont´evil, 2014b). Typi- the items in the bottom compartment would approximate
cally, the heart rate of a healthy subject displays patterns a binomial distribution.
of accelerations and deceleration at all time scales during In our terms, the bean box works the following way.
wake hours. Note, however, that the typical symmetries The items share a common history when they get into the
of multiscale fluctuations (scale symmetries) are not met funnel, and this common history leaves a trace in the re-
either. Rather, many factors impact the multiscale feature sult: depending on where the funnel is placed into the
of these variations of rhythms. For example, the current device (e.g. in the middle or not), the distribution of the
activity of the subject, her age, her life habits (smoking, items in the end varies. When the items exit the funnel,
exercising, etc.) and diseases change these multiscale fea- they take divergent paths (by bouncing on the pins) until
tures (Longo & Mont´evil, 2014b). These differences in the they reach a vertical compartment. This is, however, di-
patterns of the variability of the beat to beat interval can vergence in a weak sense. For the bean box to work, all
even be used for diagnostic purposes (West, 2006; Bailly the items have to be supposed to be symmetrical, and all
et al., 2011). the realized paths have to be supposedly taken from the
same underlying distribution. As a matter of fact, this
11
3.7. Modelization and specific objects
Current mathematical modeling practices in biology
borrow mostly the epistemology of physics and are based
on generic objects following specific trajectories. So far, we
have argued that the theorizing of physical phenomena is
based on stable mathematical structures and on the corre-
sponding analysis of generic objects. We advocate, by con-
trast, that biological organisms are specific objects moving
along possible phylo-ontogenetic trajectories. Organisms
have a historical and contextual nature and change their
organization and functions over time.
This physicomathematical modeling practice in biology
leads to many technical and epistemological problems. For
example Boolean networks (see Kauffman, 1993) are used
to model gene networks and are defined as random net-
works where the existence of an edge between two nodes
follows a given probability distribution. Such an assump-
tion is a way to model protein or gene networks in an ahis-
Figure 5: Galton’s quincux (Galton, 1894, pp.63). A ball falls but torical manner (and for example to generate them de novo
obstacles lead it to move randomly to the right or to the left. The in simulations). This disregards the fact that the actual
outcome is variability in the position of the balls at the bottom of phenomena are the result of evolution, and thus that ac-
the device. This device illustrates variation in a pre-defined set of
possibilities. Biological variation, by contrast, sometimes involves tual biological networks depend on the historical interplay
the constitution of new possibilities, which would amount for the between living beings and their environment, even at the
ball to jump outside of the quincunx. molecular level (Yamada & Bork, 2009). Hence, they are
not a sample of a random network following a given proba-
bility distribution. This is also true for cell networks: in a
assumption is necessary for the use of statistics in biol-
tissue, cell to cell interactions or the production of proteins
ogy: when performing an experiment on — say — rats,
are largely a context- and history-dependent phenomenon.
one supposes that all rats are independent realizations of
For instance, the “normalization” of a cell transferred from
a random variable taken from the same underlying distri-
a cancer tissue to a healthy one can be understood as the
bution and that this distribution is stemming from their
effect of tissue context (and its history) controlling indi-
most recent common ancestor. The most recent common
vidual cellular activities (Soto et al., 2008; Sonnenschein
ancestor plays the role of the funnel; and subsequent muta-
& Soto, 2016). These examples show that the standard
tions, effects of the environment, spontaneous variations,
modeling strategies of a biological system struggle against
etc., play the role of the pins. Variation can occur, but it
the historicity and contextuality of biological organisms.
will be merely quantitative and measured by the position
We interpret the “big data” approach, that aims at
on the horizontal axis in the bean box.
taking into account a massive amount of data in a model,
By contrast, the principle of variation posits that unex-
as an attempt to address the consequences of the histori-
pected (and unknown) qualitative variation permanently
cal nature of biological objects while keeping the physical
occurs. This means that different organisms are not differ-
methodology of establishing intelligibility on the basis of
ent realizations of a random variable taken from an under-
generic objects. Such an approach, however, raises the
lying single distribution, as this distribution cannot even
question of the intelligibility of their object, because the
be defined. In terms of the bean box, this means that the
complicated mathematical structures of models based on
pins unexpectedly open new dimensions (i.e. new relevant
big data make only computer simulations possible. Other
features arise), which would not be defined before the re-
more physical approaches focus on generic features that
alization, and would not be reproducible after either. This
even these historical systems would meet. For example,
is what we mean by divergence in a strong sense. Galton
scaling laws in networks have been extensively investi-
used his device to illustrate normal variability where vari-
gated, but their validity is criticized (Fox Keller, 2005).
ability would be quantitative, in a pre-defined space. By
Globally speaking, however, the methodological emphasis
contrast, the principle of variation implies that variation
on generic features implies that the biological meaning of
can be qualitative (i.e. symmetry changes) and that the
specific variations, and their role in a given organism, is
space of variation is not pre-defined. This, to reiterate,
lost. The issue is that without stable generic features, the
applies both at ontogenetic and phylogenetic scales18 .
this experimental methodology, which aims at selecting biological
18 Inexperimental biology, organisms are often kept as historically objects in such a way as to reduce variability “symmetrization”. A
close as possible, they may be siblings for example. The aim is more detailed analysis of biological experiments will be the object of
then to keep the divergence in their organization limited. We call a specific paper.
12
question of the objectivity of these models is open, insofar 4. Bringing organization into the picture
as their description and behavior will have a high degree
of arbitrariness: the models will miss the consequences of Let us begin with a methodological remark on the ar-
the principle of variation and, thereby, display invariants ticulation of the principles of variation and organization.
which are not valid. The theoretical definition of a biological organization at
Most mathematical models do not aim at capturing a given time is closely related to how it may change, and
features of whole organisms but, rather, at singling out that for two related reasons. First, the organization of ev-
some constraints, corresponding to specific parts of an or- ery current organism is the result of a cascade of changes
ganism. Typically, they focus on the morphogenesis of an over ontogenetic and evolutionary time scales. Second, the
organ or a tissue, for example the formation of leaf dis- appropriate theoretical definitions and representations of
position (phylotaxis) (Douady & Couder, 1996), the or- scientific objects are, generally speaking, those that enable
ganization of the cytoskeleton (Karsenti, 2008), the mor- us to understand the changes of these objects. For exam-
phogenesis of vascular networks (Lorthois & Cassot, 2010), ple, positions, momenta, and the mass are both necessary
etc. Even though this approach has obvious merits and has and sufficient to understand the changes of position of the
provided remarkable insights, it does not take into account planets of the solar system in classical mechanics. This jus-
that these organs or tissues are parts of the whole organ- tifies the theoretical representation of the planets on the
ism and that the possible reorganizations of these parts are basis of these quantities. In this respect, an appropriate
essential to variability, development, and evolution. From framework for organisms requires the articulation of orga-
a mathematical viewpoint, one aspect of this weakness can nization with the changes that it may undergo. To some
be expressed as the fact that models miss some degrees of extent, this question has been neglected in the past inso-
freedom corresponding to the changes of phase space that far as biological organization has been mostly approached
follow from changes of symmetry, in accordance with the as a mathematical fixed point, which leads to the concept
principle of variation. of organizations as maintaining themselves identically over
Although mathematical models are more and more used time.
in biology, we think that the key challenges raised by bio-
logical organisms, in particular variability, historicity and 4.1. Organization grounds constraints in specific objects
contextuality, have not yet found a proper methodological Even though organisms should be understood as spe-
and epistemological treatment. We hope that the princi- cific objects, as the principle of variation posits, we would
ples discussed in this issue will contribute to better identify argue that some of their parts exhibit generic features in
and address these challenges. a restricted sense. As mentioned in section 3.2 above, we
refer to these parts as constraints. More precisely, con-
3.8. Conclusive remarks on the principle of variation straints are characterized as entities that control biologi-
The principle of variation leads to a change of perspec- cal dynamics (processes, reactions, etc.) because of some
tive with respect to physics. Historicity, contextuality, and symmetrical (conserved) aspect which they possess at the
variability are fundamental every time an organism is un- relevant time scales. For example enzymes are not con-
der scientific scrutiny. Rather than trying to avoid the sumed in a chemical reaction that they nevertheless change
intrinsic difficulties in mathematizing these features, our completely. Similarly, the geometry of the vascular system
theoretical frame aims at building on them. To be sure, the is conserved at the time scale of blood transport, and this
randomness of symmetry changes limits the actual knowl- transport is constrained by the vascular system.
edge we can obtain on a given organism. At the same time, The stability of constraints, however, has to be ex-
however, this new kind of randomness can be studied as plained by a sound theory of biological organisms, espe-
such, and opens up new avenues of investigation. cially in the long run. Indeed, beyond the time scale at
Last, underpinning our discussion above is the fact that which a constraint operates, constraints undergo degrada-
the principle of variation involves two kinds of changes: tion. A constraint may be further stabilized by a process
changes of the biological object itself (philosophers would being under the control of another constraint, which is it-
say this is an ontological change) and changes in the ques- self stabilized by another constrained process, and so on:
tion asked about this object (philosophers would say this if the chain of dependencies folds up and the constraints
is an epistemological change). For example, developmental can be said to be mutually dependent, the system of con-
biology studies features that appeared with multi-cellularity: straints is organized. The constraints that constitute an
the field is thus a result of biological variation. Recipro- organism are the organized ones, which (i) act on a process
cally, growing cells in lab conditions comes with modifica- stabilizing a constraint of the organism and (ii) depend on
tions of their behaviors which in turns affects the questions at least another constraint of the organism. The key as-
at stake and possibly their future culture conditions. Thus, pect in this framework is that constraints are stable at a
in our view, the instability of biological objects goes hand given time-scale, while being stabilized by processes tak-
in hand with the instability of biological questions: they ing place at other time scales, so that constraints behave
co-constitute each other. as local invariants with respect to the processes they con-
strain.
13
However, while the time scales of constraints in the In a theoretical sense, the generality of biological anal-
principle of organization are the intrinsic time scales of ysis is made possible by the permanent relevance of orga-
the processes and constraints under study, they do not nization as closure, despite the continual symmetry and
preclude a change at these or other time scales for reasons phase space changes. To a certain extent, the situation
extrinsic to these objects. Changes of organization stem- of closure is similar to that of the energy of a physical
ming from the principle of variation can alter a constraint system being conserved despite its permanent changes of
at any time scale. In this event, the former constraint may state. In the case of a change of constraints, an organized
lose its status of constraint or may operate differently. object goes from one closed regime to another, unless the
The cohesion of organisms is one of their fundamental organism does not succeed in establishing a new regime
features, and this cohesion has been the object of many and dies.
theoretical investigations, for example as autopoiesis (Varela
et al., 1974) or as work-constraints cycles (Kauffman, 2002). 4.3. Relevant variation with respect to closure
Following this line, we argue that biological organisms real- The principle of organization understood as the closure
ize closure of constraints (Mont´evil & Mossio, 2015; Mossio of constraints leads to the idea that the relevant changes
et al., 2016): functional parts of organisms act as con- of the organism involve constraints subject to closure. The
straints on each other, and they realize a mutual depen- changes of constraints that do not impact the constraints
dence. Closure is basically the circularity in the relation subject to closure fall in two categories: those that af-
of dependence between constraints. The principle of or- fect the environment and those that affect the organism
ganization that we propose states that the constraints of (in other aspects than the constraints subject to closure).
organisms realize closure. If a change of constraint affects the environment, it may
We postulate that the stability of functional constraints be biologically relevant, for example if it affects other or-
hinges on their mutual dependence (Mont´evil & Mossio, ganisms. If the change affect the organism, but not its
2015; Mossio et al., 2016), so that the overall stability organized constraints, then it is not significant for the or-
of biological organisms is justified by the closure of con- ganism in the light of the principle of organization: these
straints. When we consider the principle of organization constraints do not play a role in the biological system (al-
and the principle of variation together, constraints are con- though they may be involved in an unpredictable organi-
tingent in two complementary ways. They are contingent zational change).
because of their historical nature and because their exis- As for the changes that may affect the organization, a
tence depends on the circularity of closure instead of being general distinction can be made between irrelevant and
grounded on other stable first principles. functional variations. On the one hand, processes and
constraints may undergo irrelevant variations, for exam-
4.2. The epistemological status of closure under variation ple small quantitative variations, i.e. quantitative fluctua-
By relying on the principle of organization, it is the- tions that neither undermine nor modify the overall orga-
oretically meaningful to work on sets of constraints that nization. This is variation that, in a word, the organism
verify closure. Following the principle of variation, how- does not need to control in order to ensure its stability,
ever, constraints are not necessarily conserved over time by hypothesis. On the other hand, variation can be func-
and may undergo changes which cannot be stated in ad- tional, in the sense of resulting in a change of one of more
vance. As a result, the validity of closure must extend constraints, of their relations, and hence of the very organi-
beyond a given configuration of constraints. The validity zation. Examples of quantitative variation are for example
of the principle of organization should not be understood moderate differences in the weight of some organs like the
as based on a given set of constraints (or invariants) which liver, or in enzyme concentrations; examples of functional
would happen to realize closure. Accordingly, the princi- variations are the reshaping of bones and musculature to
ple of organization is not deducible from a set of invariants perform a new function or to perform differently an old
and symmetries (as in “physical laws”), rather, it is the function (West-Eberhard, 2003). Of course, the quantita-
condition of possibility for the existence and persistence tive variation of a given constraint can also be potentially
of biological constraints (i.e. local invariants and symme- functional, in the sense of enabling the possible further
tries). For this reason, we suggest that closure constitutes emergence of functional variation, including pathological
the principle of organization that, alongside variation and ones.
other principles, frames the biological domain as a whole Another example of functional variation is random
(see Mont´evil & Mossio, 2015; Mossio et al., 2016). gene expression, which has been studied both in unicel-
In epistemological terms, stating that the principle of lular (Eldar & Elowitz, 2010) and multicellular organisms
organization is a fundamental principle implies that it can- (Dueck et al., 2016). In this literature however, functional
not be deduced from underlying stable symmetries and be- variation is mostly understood in an evolutionary sense,
comes an (irreducible) theoretical principle for biological while closure provides a systemic interpretation of func-
organisms. Closure becomes an a priori that replaces the tions (Mossio et al., 2009). As a result, closure enables
a priori of space and time in physics, or, more precisely, us to conceive functional variation that is not necessar-
of the phase spaces of physical theories.
14
ily inherited, provided that the constraint resulting from cells by albumin: after a cell division (which is a
variation is still subject to closure. symmetry change19 ), the same albumin maintains
its inhibitory effect on the new cells.
4.4. Closure remains closure under variation
• Another situation corresponds to a change involv-
As discussed in Mossio et al. (2016), closure contributes
ing a modification of the relationship between pre-
to making both internal and external variations possible.
existing constraints as they come together to gener-
The circularity of closure weakens the coupling between
ate a new biologically structure or dynamics. Such
what is going on inside a system and its boundary condi-
a change is fundamentally non-local with respect to
tions (Barandiaran & Moreno, 2008). Such a decoupling
the graph of interacting constraints. In this kind of
enables variation beyond what would be permissible if the
situation, some constraints act on processes which
system were completely determined by its boundary condi-
they did not constrain before the change. This cor-
tions (such as in physical self-organization). Reciprocally,
responds typically to the notion of exaptation. In
an organism can stand a relatively unstable ecosystem be-
general, such a change implies the alteration or the
cause of its autonomous stability due to the closure of con-
appearance of specific constraints that establish the
straints.
new behavior: the important difference with respect
Under our principles, functional variation cannot lead
to the case described in the previous paragraph is
to a violation of the organization principle — except in the
that various other constraints are also mandatory to
case of death. This means that any change affecting the
enable the emergence of the new behavior.
constitutive constraints are changes from one organized
situation to another. In our frame, closure is always met, • Finally, a change in organization might result from
even though the constraints relevant to closure may and the appearance of new constraints. In order for a new
do change. The continuous alteration, loss or acquisition constraint to be included in the closed system, the
of functions result in the realization of new organizational organization must be reshaped so that the new con-
regimes; each regime, in turn, achieves a form of stability straint be integrated to the organization (Mont´evil
determined by closure as a mutual stabilization of con- & Mossio, 2015; Mossio et al., 2016). There are two
straints. Being subject to both the organization and vari- aspects to this: the new constraint must be main-
ation principles, biological organisms realize a succession tained by other constraints (I) and maintain another
of different instances of organized regimes over individual constraint (II). Whether (I) or (II) occurs first cor-
and evolutionary times. Then, the stability achieved by responds to different scenarios. It is fairly easy to
the organism is not conservative, but it is for a part cumu- picture a constraint being maintained (criterion I)
lative, insofar as it sustains many functional innovations, starting to play a role in an organization after some
and makes their preservation over time possible. time (criterion II). For instance, in mammalian de-
Changes of the organization may correspond to several velopment, lungs are first formed and maintained (I)
situations depending on the constraints involved. They and they acquire a functional role only after birth
may be more or less local with respect to the rest of the (II). However, the opposite may also happen, for ex-
organization. We propose a typology on this basis: ample, thanks to generic physiological responses dis-
cussed above: a change of behavior leading to me-
• A first situation consists of a local change of a con- chanical friction (II) leads to the strengthening of the
straint, such that it does not induce a change in skin by keratinization (I). Lastly, the two aspects can
the relationship between constraints. For example, a be coupled. For instance, some structures (such as
supplementary branching event in a network or tree muscles, bones, etc.) which are not used (II) may
structure (such as vascular networks or mammary atrophy (I), and reciprocally their use (II) may lead
glands) does not correspond to a major reorganiza- to their further development and strengthening (I).
tion of the constraints of a system. Let us remark,
still, that this situation corresponds to a basic sym- The key issue about changes of organization is the in-
metry breaking involving the appearance of new rele- scription of the change in a new organization. After a
vant quantities of preexisting kinds (for example the change of constraints takes place at relatively short time
angle between the new branches). Therefore, such a scales, the altered constraints involved may be stabilized
change is generic (a branching among many possible by other constraints, at longer time scales. These sta-
branchings). In section 3.1 and 3.3 above, we dis- bilizing constraints are then typically solicited differently
cussed such examples of generic symmetry changes than before the change: they maintain, for example, the
in the context of specific objects. In turn, the new
constraints can be stabilized by generic constraints
19 Cell division corresponds to the disappearance of an object and
(insofar as the new branch is stabilized in the same
manner than the preexisting ones). In the context of the appearance of two new non-identical objects, see Sonnenschein
& Soto (1999; Longo et al. (2015; Soto et al. (2016; Mont´ evil et al.
closure, a simple example of a generic stabilization is (2016).
the inhibition of the proliferation of estrogen-target
15
same tissues but in a different macroscopic shape or con- a “organismal specification” of the framing principle. The
figuration. This can happen through generic physiological latter is an informal overarching principle that can be fur-
responses (e.g. keratinization of oral mucosa subjected ther specified by the two principles of organization and
to friction, resorption of bones under compressive stress, variation.
etc.). These changes do not happen only in the interac- The framing principle applies to morphogenesis under-
tion with the environment, they happen in essential devel- stood in a general sense, that is, both to organogenesis
opmental, metabolic and regulatory processes (as in the and to proliferation with variation at the cellular level.
developmental processes mentioned above). Another ex- In other words, both in organ generation (for example,
ample is given in David et al. (2013): the authors show lungs, vascular systems, plants’ organs etc.) and in repro-
that “jamming” the regulation of key metabolic genes of duction, a form is iteratively (and hereditarily) produced,
yeast cells did not lead to their death but, instead, to new yet never identically. Let us now develop what it means
dynamic behaviors which enabled them to thrive after a for biological phenomena to be “non-identical iterations of
transition period. morphogenetic processes”.
A change of constraint, or the appearance of a new con- By non-identical, we mean (as discussed above) not
straint does not necessarily lead to a stabilization of the just quantitative changes but rather unpredictable changes
new situation. In particular, organized constraints might of symmetry, thus unpredictable qualitative changes in the
tend to restore the initial situation because constraints behavior of the object. In the context of the organism, the
subject to closure are maintained by another constrained relevant changes are the ones impacting the organization,
process. For example, a mutation in mrna is not going that is to say, the ones changing the constraints subject to
to be sustained because the production of new mrna will closure.
not carry the same variation. One might refer to such a The iterations are those of organized objects, subject
tendency as a form of organizational “inertia”. In such a to closure. However, they should be understood in several
case, the new constraint may vanish at a relatively short ways depending on the particular kind of objects they refer
time scale. The diametrically opposite situation (among to.
others) is also possible. It corresponds to an amplification First, closure is by definition about circular causal ar-
of a change affecting a constraint, which in turn destabi- chitectures. For instance, consider a simple closed system,
lizes other constraints in the longer run. It is typically the where C1 generates C2 (at time-scale τ1 ), C2 generates C3
case in carcinoma where, as stated by the Tissue Field Or- (at time-scale τ2 ), and C3 generates C1 (at time-scale τ3 ,
ganization Theory of carcinogenesis, the lack of sufficient say this is the fastest of the three). To discuss iterations,
constraints on the epithelium can lead to a progressive let us consider a perturbation on C1 at t0 . This pertur-
disorganization of the tissue and, sometimes, disrupt the bation impacts C2 significantly at time t0 + τ2 . Then, C2
organization of the whole organism leading to death (Son- impacts C3 at time t0 + τ2 + τ3 . Finally C3 impacts C1 at
nenschein & Soto, 2016). time t0 +τ2 +τ3 +τ1 , and this closes the loop.20 Thus, with
Overall, the principle of variation complements the prin- the flow time, the circularities of closure lead to iterations
ciple of organization, which should not be conceived as a of closed patterns. More generally, in a loop described by
“fixed point” that iterates itself always identically. Rather, closure, the duration of the loop as a whole corresponds
organisms change while staying organized. Variation par- to the scale of the slowest process. At this time scale, the
ticipates in the robustness of closure in changing envi- iterations are the whole set of constrained processes which
ronments. Changes of organization actually enabled the stabilize and maintain the organization of the organism.
maintenance of organisms over very long time scales (dur- Following the principle of variation, these iterations are
ing evolution). Last, but not least, current organisms are associated with unpredictable changes of symmetry.
the product of such variations. Current biological orga- Second, the organizations themselves are iterated. This
nizations are determined by their (partially) cumulative adds to the principle of organization the notion of repro-
variations, and this process enables organisms to explore duction. By reproduction we mean the process of go-
more and more complex organization (Gould, 1997; Bailly ing from one organized object to two (or more) organized
& Longo, 2009; Longo & Mont´evil, 2012; Soto et al., 2016). objects21 . Reproduction pertains to the notion that the
default state of cells is proliferation (with variation and
5. Non-identical iteration of morphogenetic pro- motility) (Soto et al., 2016; Longo et al., 2015) which com-
cesses plements the principles of organization and variation.
Reproduction is also essential in that organizations which
As a last step, we discuss in this section the connection undergo variations may undergo deleterious variations. As
between the organization and variation principles and the
“framing principle” proposed in Longo et al. (2015), ac- 20 Note that the iterations of these loops are not just about suc-
cording to which biological phenomena should be under- cessive operations. Instead, all constraints are active simultaneously.
Incidentally, this is why a perturbation approach is better suited to
stood as “non-identical iterations of morphogenetic pro- show the iterative structure underlying closure.
cesses”. As mentioned in Mossio et al. (2016), we submit 21 Note that some situations can be fairly complex. Indeed, some
that organization and variation, taken together, constitute organizations include constraints which act across generations.
16
a thought experiment, a cell which would never prolifer- without stressing the importance of variation, its perva-
ate but would undergo variation should have a finite life siveness and its conceptual consequences. This has led
expectancy because at some point a deleterious variation modeling attempts to focus on generic objects, which are,
would occur. As a result, varying organizations require we think, unable to adequately represent current biological
reproduction to be sustained in an open-ended manner. objects.
Reproduction enables a balance between the exploration In order to understand current biological objects, ar-
of possibly morbid variations and the maintenance of a ticulating the principle of variation with the principle of
strain of organized systems. organization is necessary. In our framework, organization
Finally, the framing principle applies also to organ for- grounds the relative stability of a set of constraints by the
mation. Iteration is a very common morphogenetic process circularity of closure. It controls and counters (a part of
which takes place for example in branching morphogenesis the) variation that would be deleterious and would un-
of glands such as the mammary glands, the lungs, etc22 . dermine the very existence of the organism. At the same
Iteration processes explain the abundance of fractal-like time, organisms undergo quantitative and functional vari-
structures in biology (Longo & Mont´evil, 2014b). Such ations, both of them being crucial requirements for their
multi-scale structures play a particular role because they increase in complexity, their adaptability, and, in the end,
link different scales, coupling macroscopic and microscopic the sustainability of organization itself as suggested in
entities. As such they constitute spatio-temporal coher- Ruiz-Mirazo et al. (2004). One of the central challenges
ence structures, which we propose elsewhere to interpret of a full-fledged theory of organisms consists in providing
as biological levels of organization (Longo et al., 2012b). a coherent account of how they manage simultaneously to
restrict and undergo variation.
The epistemological structure of our framework is dis-
6. Conclusions: back to theoretical principles
tinct from the one of physical theories. In physical the-
Biological variation is relevant at all levels of organiza- ories, assumptions on the validity of stable mathematical
tion, and, for example, it is manifested in the default state structures (symmetries) come first, and they may lead to
of cells (proliferation with variation and motility). The randomness in a given mathematical space. In our frame-
principle of variation that we propose states that biologi- work, variability comes first and closure justifies the valid-
cal organisms are specific objects and, thereby, fundamen- ity of constraints.
tally different from the objects defined in physical theories. The notion of constraint is central to our framework.
The principle, which draws directly on Darwin’s insights Constraints are the building blocks of mathematical mod-
on biological variation, embeds a specific notion of ran- eling in biology and are the main objects of experimental
domness, which corresponds to unpredictable changes of investigation. The theoretical notion of constraints that
the mathematical structure required to describe biological we propose should lead to a reinterpretation of mathemat-
objects. In this framework, biological objects are inher- ical models that are based on them. In our framework
ently variable, historical and contextual. A specific object constraints depend on the rest of the organism and the
such as an organism is fundamentally defined by its his- rest of the organism depends on them (principle of organi-
tory and context. Its constraints which may be described zation). Moreover, constraints may undergo unpredictable
by mathematical structures are the result of a history and variations (principle of variation).
may change over time. The principles of variation and organization do not aim
Our approach to variation contrasts with a relevant at providing a complete framework to understand biologi-
part of the theoretical literature on biological organization cal objects (the default state, for instance, is also required),
which aims at investigating the origin of life by the means but they elaborate on both the Darwinian and organi-
of minimal or physical models. The strong point of these cist traditions and lead to a significant departure from the
models is that they lead to tractable mathematics (see for physical methodology, which opens the way to original re-
example Luisi, 2003; Ruiz-Mirazo & Moreno, 2004). Here, search directions.
we aim instead at combining organization and variation in
a framework that focuses on current organisms, with the Acknowledgments
massive amount of history that they carry. This difference
between the two methodologies corresponds to distinct but Ma¨el Mont´evil’s work is supported by the ‘Who am
complementary aims, and, crucially, to the fact that the I’ Labex grant (ANR-11-LABX-0071 WHOAMI). Arnaud
concept of organization has been traditionally approached Pocheville has been partly supported by Ana Soto’s Blaise
´
Pascal Chair, Ecole Normale Sup´erieure, R´egion ˆIle-de-
France. The authors are grateful to Cheryl Schaeberle for
22 Note that iterations in organ formation are not just iterations
her linguistic corrections and to Ana Soto for her editorial
of a shape (such as iterations of branching): they involve the whole
set of constraints which enable the maintenance of shape. In the work.
case of epithelial branching structures for instance, this includes the
basement membrane and the activity of stromal cell which maintains
this membrane and the collagen of the tissue around a new branch.
17
References Gould, S. J., & Vrba, E. S. (1982). Exaptation — a missing term in
the science of form. Paleobiology, 8 , 4–15.
Bailly, F., & Longo, G. (2009). Biological organization and anti- Griffiths, P. E., Pocheville, A., Calcott, B., Stotz, K., Kim, H., &
entropy. Journal of Biological Systems, 17 , 63–96. doi:10.1142/ Knight, R. (2015). Measuring causal specificity, .
S0218339009002715. Karsenti, E. (2008). Self-organization in cell biology: a brief history.
Bailly, F., & Longo, G. (2011). Mathematics and the natural sci- Nature Reviews Molecular Cell Biology, 9 , 255–262. doi:10.1038/
ences; The Physical Singularity of Life. London: Imperial Col- nrm2357.
lege Press. Preliminary version in French: Hermann, Vision des Kauffman, S. (2002). Investigations. Oxford University Press, USA.
sciences, 2006. Kauffman, S. A. (1993). The origins of order: Self organization and
Bailly, F., Longo, G., & Mont´ evil, M. (2011). A 2-dimensional ge- selection in evolution. Oxford university press.
ometry for biological time. Progress in Biophysics and Molecular Kauffman, S. A. (2013). Evolution beyond newton, darwin, and en-
Biology, 106 , 474 – 484. doi:10.1016/j.pbiomolbio.2011.02.001. tailing law. Beyond Mechanism: Putting Life Back Into Biology,
Barandiaran, X., & Moreno, A. (2008). Adaptivity: From (p. 1).
metabolism to behavior. Adaptive Behavior , 16 , 325–344. Longo, G., & Mont´ evil, M. (2011). From physics to biology by
Beatty, J. (1995). The evolutionary contingency thesis. In Concep- extending criticality and symmetry breakings. Progress in Bio-
tual Issues in Evolutionary Biology. (p. 217). Cambrdige, MA: physics and Molecular Biology, 106 , 340 – 347. doi:10.1016/j.
MIT Press. pbiomolbio.2011.03.005. Invited paper, special issue: Systems
Begley, C. G., & Ellis, L. M. (2012). Drug development: Raise Biology and Cancer.
standards for preclinical cancer research. Nature, 483 , 531–533. Longo, G., & Mont´ evil, M. (2012). Randomness increases or-
URL: http://dx.doi.org/10.1038/483531a. der in biological evolution. In M. Dinneen, B. Khoussainov, &
Bowler, P. J. (2005). Variation from darwin to the modern synthesis. A. Nies (Eds.), Computation, Physics and Beyond (pp. 289 –
In B. Hallgr´ımsson, & B. K. Hall (Eds.), Variation: A Central 308). Springer Berlin / Heidelberg volume 7160 of Lecture Notes
Concept in Biology. Burlington, MA, USA: Elsevier. in Computer Science. doi:10.1007/978-3-642-27654-5\_22 in-
Brisken, C., Socolovsky, M., Lodish, H. F., & Weinberg, R. (2002). vited paper, Auckland, New Zealand, February 21-24, 2012.
The signaling domain of the erythropoietin receptor rescues pro- Longo, G., & Mont´ evil, M. (2013). Extended criticality, phase spaces
lactin receptor-mutant mammary epithelium. Proceedings of the and enablement in biology. Chaos, Solitons & Fractals, (pp. –).
National Academy of Sciences, 99 , 14241–14245. doi:10.1073/ doi:10.1016/j.chaos.2013.03.008.
pnas.222549599. Longo, G., & Mont´ evil, M. (2014a). Perspectives on Organ-
Byers, N. (1999). E. Noether’s discovery of the deep connection isms: Biological time, symmetries and singularities. Lecture
between symmetries and conservation laws. In The heritage of Notes in Morphogenesis. Dordrecht: Springer. doi:10.1007/
Emmy Noether in algebra, geometry, and physics (pp. 67–81). 978-3-642-35938-5.
Tel Aviv: Israel mathematical conference volume 12. Longo, G., & Mont´ evil, M. (2014b). Scaling and scale symmetries in
Collective (2005). Variation: A Central Concept in Biology. Burling- biological systems. In Perspectives on Organisms Lecture Notes
ton, MA, USA: Elsevier. in Morphogenesis (pp. 23–73). Springer Berlin Heidelberg. doi:10.
Danchin, E., & Pocheville, A. (2014). Inheritance is where physiol- 1007/978-3-642-35938-5_2.
ogy meets evolution. The Journal of Physiology, 592 , 2307–2317. Longo, G., & Mont´ evil, M. (2014c). Symmetry and symmetry break-
doi:10.1113/jphysiol.2014.272096. ings in physics. In Perspectives on Organisms Lecture Notes
Darwin, C. (1859). On the Origin of Species by Means of Natural in Morphogenesis (pp. 121–136). Springer Berlin Heidelberg.
Selection, or the Preservation of Favoured Races in the Struggle doi:10.1007/978-3-642-35938-5_5.
for Life. London: John Murray. Longo, G., & Mont´ evil, M. (to appear). Comparing symmetries
David, L., Ben-Harosh, Y., Stolovicki, E., Moore, L. S., Nguyen, M., in models and simulations. In M. Dorato, L. Magnani, &
Tamse, R., Dean, J., Mancera, E., Steinmetz, L. M., & Braun, E. T. Bertolotti (Eds.), Springer Handbook of Model-Based Science.
(2013). Multiple genomic changes associated with reorganization Springer.
of gene regulation and adaptation in yeast. Molecular Biology and Longo, G., Mont´ evil, M., & Kauffman, S. (2012a). No entailing laws,
Evolution, 30 , 1514–1526. doi:10.1093/molbev/mst071. but enablement in the evolution of the biosphere. In Genetic and
De Klerk, G. (1979). Mechanism and vitalism. a history of the con- Evolutionary Computation Conference. GECCO’12 New York,
troversy. Acta Biotheoretica, 28 , 1–10. doi:10.1007/BF00054676. NY, USA: ACM. doi:10.1145/2330784.2330946 invited Paper.
Douady, S., & Couder, Y. (1996). Phyllotaxis as a dynamical self Longo, G., Mont´ evil, M., & Pocheville, A. (2012b). From bottom-up
organizing process part i: The spiral modes resulting from time- approaches to levels of organization and extended critical transi-
periodic iterations. Journal of Theoretical Biology, 178 , 255 – tions. Frontiers in Physiology, 3 . doi:10.3389/fphys.2012.00232.
273. doi:dx.doi.org/10.1006/jtbi.1996.0024. Invited paper.
Dueck, H., Eberwine, J., & Kim, J. (2016). Variation is function: Longo, G., Mont´ evil, M., Sonnenschein, C., & Soto, A. M. (2015).
Are single cell differences functionally important? BioEssays, 38 , In search of principles for a theory of organisms. Journal of Bio-
172–180. doi:10.1002/bies.201500124. sciences, (pp. 1–14). doi:10.1007/s12038-015-9574-9.
Eldar, A., & Elowitz, M. B. (2010). Functional roles for noise in Longo, G., Palamidessi, C., & Paul, T. (2011). Some bridging re-
genetic circuits. Nature, 467 , 167–173. sults and challenges in classical, quantum and computational ran-
Elowitz, M. B., Levine, A. J., Siggia, E. D., & Swain, P. S. (2002). domness. In Randomness Through Computation: Some Answers,
Stochastic gene expression in a single cell. Science, 297 , 1183 – More Questions (pp. 73–91). World Scientific.
1186. doi:10.1126/science.1070919. Lorthois, S., & Cassot, F. (2010). Fractal analysis of vascular net-
Fox Keller, E. (2005). Revisiting” scale-free” networks. BioEssays, works: Insights from morphogenesis. Journal of Theoretical Biol-
27 , 1060–1068. doi:10.1002/bies.20294. ogy, 262 , 614–633. doi:10.1016/j.jtbi.2009.10.037.
Galton, F. (1894). Natural Inheritance.. Macmillan. Luisi, P. L. (2003). Autopoiesis: a review and a reappraisal. Natur-
Gilbert, S. F., & Epel, D. (2009). Ecological developmental biology: wissenschaften, 90 , 49–59.
integrating epigenetics, medicine, and evolution. Sinauer Asso- Miquel, P.-A., & Hwang, S.-Y. (2016). From physical to biological
ciates Sunderland. individuation. Progress in biophysics and Molecular Biology, .
Gould, S. J. (1989). Wonderful life: The Burgess Shale and the doi:10.1016/j.pbiomolbio.2016.07.002. This issue.
Nature of History.. New York, USA: Norton. Mont´ evil, M., & Mossio, M. (2015). Biological organisation as closure
Gould, S. J. (1997). Full house: The spread of excellence from Plato of constraints. Journal of Theoretical Biology, 372 , 179 – 191.
to Darwin. Three Rivers Pr. doi:dx.doi.org/10.1016/j.jtbi.2015.02.029.
Gould, S. J. (2002). The structure of evolutionary theory. Harvard Mont´ evil, M., Speroni, L., Sonnenschein, C., & Soto, A. (2016). Mod-
University Press. eling mammary organogenesis from biological first principles: cells
18
and their physical constraints. Progress in biophysics and Molecu- West, B. (2006). Where medicine went wrong: Rediscovering the
lar Biology, . doi:10.1016/j.pbiomolbio.2016.08.004. This issue. path to complexity volume Volume 11 of Studies of nonlinear phe-
Moreno, A., & Mossio, M. (2015). Biological autonomy. A Philo- nomena in life sciences. Teaneck, NJ: World Scientific.
sophical and Theoretical Enquiry.. Dordrecht: Springer. West, G., & Brown, J. (2005). The origin of allometric scaling laws in
Mossio, M., Mont´ evil, M., & Longo, G. (2016). Theoretical principles biology from genomes to ecosystems: Towards a quantitative uni-
for biology: Organization. Progress in biophysics and Molecular fying theory of biological structure and organization. Journal of
Biology, . doi:10.1016/j.pbiomolbio.2016.07.005. This issue. Experimental Biology, 208 , 1575–1592. doi:10.1242/jeb.01589.
Mossio, M., Saborido, C., & Moreno, A. (2009). An organizational West-Eberhard, M. J. (2003). Developmental plasticity and evolu-
account of biological functions. The British Journal for the Phi- tion. Oxford University Press.
losophy of Science, 60 , 813–841. Yamada, T., & Bork, P. (2009). Evolution of biomolecular net-
Pocheville, A., & Danchin, E. (2016). Genetic assimilation and the works—lessons from metabolic and protein interactions. Na-
paradox of blind variation. In D. Walsh, & P. Huneman (Eds.), ture Reviews Molecular Cell Biology, 10 , 791–803. doi:10.1038/
Challenges to Evolutionary Theory. Oxford: Oxford University nrm2787.
Press.
Pross, A., & Pascal, R. (2013). The origin of life: what we know,
what we can know and what we will never know. Open Biology,
3 . doi:10.1098/rsob.120190.
Rivenbark, A. G., O’Connor, S. M., & Coleman, W. B. (2013).
Molecular and cellular heterogeneity in breast cancer. The Amer-
ican Journal of Pathology, 183 , 1113–1124.
Ruiz-Mirazo, K., & Moreno, A. (2004). Basic autonomy as a funda-
mental step in the synthesis of life. Artificial life, 10 , 235–259.
Ruiz-Mirazo, K., Peret´ o, J., & Moreno, A. (2004). A universal def-
inition of life: Autonomy and open-ended evolution. Origins of
life and evolution of the biosphere, 34 , 323–346. doi:10.1023/B:
ORIG.0000016440.53346.dc.
Socolovsky, M., Dusanter-Fourt, I., & Lodish, H. F. (1997). The
prolactin receptor and severely truncated erythropoietin recep-
tors support differentiation of erythroid progenitors. Journal of
Biological Chemistry, 272 , 14009–14012. doi:10.1074/jbc.272.
22.14009.
Sonnenschein, C., & Soto, A. (1999). The society of cells: cancer
and control of cell proliferation. New York: Springer Verlag.
Sonnenschein, C., & Soto, A. (2016). Carcinogenesis explained
within the context of a theory of organisms, progress in biophysics
and molecular biology. Progress in biophysics and Molecular Bi-
ology, . doi:10.1016/j.pbiomolbio.2016.07.004. This issue.
Soto, A., & Longo, G. (2016). Why do we need theories? Progress in
biophysics and Molecular Biology, . doi:10.1016/j.pbiomolbio.
2016.06.005. This issue.
Soto, A., Longo, G., Mont´ evil, M., & Sonnenschein, C. (2016).
The biological default state of cell proliferation with variation
and motility, a fundamental principle for a theory of organisms.
Progress in biophysics and Molecular Biology, . doi:10.1016/j.
pbiomolbio.2016.06.006. This issue.
Soto, A., Sonnenschein, C., & Miquel, P.-A. (2008). On phys-
icalism and downward causation in developmental and can-
cer biology. Acta Biotheoretica, 56 , 257–274. doi:10.1007/
s10441-008-9052-y.
Soto, A. M., & Sonnenschein, C. (2005). Emergentism as a de-
fault: Cancer as a problem of tissue organization. Journal of
Biosciences, 30 , 103–118. doi:10.1007/BF02705155.
Stoer, J., & Bulirsch, R. (2013). Introduction to numerical analysis
volume 12. Springer Science & Business Media.
Thomas-Vaslin, V., Six, A., Bellier, B., & Klatzmann, D. (2013).
Encyclopedia of systems biology. chapter Lymphocyte Dynam-
ics and Repertoires, Modeling. (pp. 1149–1152). New York, NY:
Springer New York. doi:10.1007/978-1-4419-9863-7_96.
Turing, A. M. (1952). The chemical basis of morphogenesis. Philo-
sophical Transactions of the Royal Society of London. Series B,
Biological Sciences, 237 , 37–72. doi:10.1098/rstb.1952.0012.
Van Fraassen, B. (1989). Laws and symmetry. Oxford University
Press, USA.
Varela, F., Maturana, H., & Uribe, R. (1974). Autopoiesis: The
organization of living systems, its characterization and a model.
Biosystems, 5 , 187 – 196. doi:10.1016/0303-2647(74)90031-8.
Vogel, D., Nicolis, S. C., Perez-Escudero, A., Nanjundiah, V.,
Sumpter, D. J. T., & Dussutour, A. (2015). Phenotypic vari-
ability in unicellular organisms: from calcium signalling to social
behaviour. Proceedings of the Royal Society of London B: Biolog-
ical Sciences, 282 . doi:10.1098/rspb.2015.2322.
19
READ PAPER