BioSystems 69 (2003) 95–114
Cooperation and conflict in the evolution of individuality
IV. Conflict mediation and evolvability in Volvox carteri
Richard E. Michod∗ , Aurora M. Nedelcu, Denis Roze
Department of Ecology and Evolutionary Biology, University of Arizona Tucson, AZ 85721, USA
Abstract
The continued well being of evolutionary individuals (units of selection and evolution) depends upon their evolvability, that
is their capacity to generate and evolve adaptations at their level of organization, as well as their longer term capacity for
diversifying into more complex evolutionary forms. During a transition from a lower- to higher-level individual, such as the
transition between unicellular and multicellular organisms, the evolvability of the lower-level (cells) must be restricted, while
the evolvability of the new higher-level unit (multicellular organism) must be enhanced. For these reasons, understanding the
factors leading to an evolutionary transition should help us to understand the factors underlying the emergence of evolvability of
a new evolutionary unit. Cooperation among lower-level units is fundamental to the origin of new functions in the higher-level
unit. Cooperation can produce a new more complex evolutionary unit, with the requisite properties of heritable fitness variations,
because cooperation trades fitness from a lower-level (the costs of cooperation) to the higher-level (the benefits for the group).
For this reason, the evolution of cooperative interactions helps us to understand the origin of new and higher-levels of fitness
and organization. As cooperation creates a new level of fitness, it also creates the opportunity for conflict between levels of
selection, as deleterious mutants with differing effects at the two levels arise and spread. This conflict can interfere with the
evolvability of the higher-level unit, since the lower and higher-levels of selection will often “disagree” on what adaptations are
most beneficial to their respective interests. Mediation of this conflict is essential to the emergence of the new evolutionary unit
and to its continued evolvability. As an example, we consider the transition from unicellular to multicellular organisms and study
the evolution of an early-sequestered germ-line in terms of its role in mediating conflict between the two levels of selection, the
cell and the cell group. We apply our theoretical framework to the evolution of germ/soma differentiation in the green algal group
Volvocales. In the most complex member of the group, Volvox carteri, the potential conflicts among lower-level cells as to the
“right” to reproduce the higher-level individual (i.e. the colony) have been mediated by restricting immortality and totipotency to
the germ-line. However, this mediation, and the evolution of an early segregated germ-line, was achieved by suppressing mitotic
and differentiation capabilities in all post-embryonic cells. By handicapping the soma in this way, individuality is ensured, but
the solution has affected the long-term evolvability of this lineage. We think that although conflict mediation is pivotal to the
emergence of individuality at the higher-level, the way in which the mediation is achieved can greatly affect the longer-term
evolvability of the lineage.
© 2002 Elsevier Science Ireland Ltd. All rights reserved.
Keywords: Levels of selection; Germ-line; Mutation; Green algae; Altruism; Group selection
1. Evolvability, evolutionary transitions and
fitness
∗ Corresponding author. Tel.: +1-520-621-7517/7509;
fax: +1-520-621-9190.
E-mail address: michod@u.arizona.edu (R.E. Michod).
The continued well being of evolutionary individuals depends upon their evolvability, which, in our
0303-2647/02/$ – see front matter © 2002 Elsevier Science Ireland Ltd. All rights reserved.
PII: S 0 3 0 3 - 2 6 4 7 ( 0 2 ) 0 0 1 3 3 - 8
96
R.E. Michod et al. / BioSystems 69 (2003) 95–114
view, includes not only their capacity to generate and
evolve adaptations, but also as their capacity to diversify into more complex evolutionary units. To adapt
evolutionary individuals must be units of selection,
and, ever since Darwin, we have understood that a unit
of selection must possess the following properties: the
struggle to survive; variation; and, heritability, so that
offspring resemble parents. We may combine these
three properties and simply say that a unit of selection
must have heritable variation in fitness. There are a
variety of such units in biology: genes, gene networks,
prokaryote cells, eukaryotic cells (cells in cells), multicellular organisms, and groups and societies. How
and why did these different kinds of evolutionary
units arise?
In the present paper, we consider our theory for
the origin and evolution of new kinds of individuals
(Michod, 1996, 1997, 1999; Michod and Roze, 1997,
1999, 2000) from the perspective of evolvability.
During a transition to a higher-level individual, such
as between unicellular and multicellular organisms,
the evolvability of lower-level units (for example,
cells) must be restricted, while the evolvability of
the new higher-level unit (for example, the multicellular organism) must be enhanced. Consequently,
understanding the factors leading to an evolutionary
transition should help us to understand the factors
underlying evolvability.
We study the role of the process of conflict mediation in enhancing the evolvability of a new unit of
selection. As an application of our theory, we consider
a general feature of multicellular organisms, the separation and specialization of germ and somatic-lines
of cells, and apply our results to the evolution of
germ/soma differentiation in the green algal group,
Volvocales (Nedelcu and Michod, 2003). We use the
terms “germ” and “soma” in the sense of there being
two kinds of cells in the multicellular group, cells
that are specialized in contributing to the next generation of individuals and cells that are specialized
in vegetative functions and do not directly reproduce
the next generation of individuals. Even organisms
often regarded as not having a germ-line, such as
plants, have cells that are specialized in reproductive
and vegetative functions, and so meet our criteria of
reproductive specialization. Specialization of cells in
reproductive and vegetative functions is an almost
universal feature of multicellular life.
The basic problem in an evolutionary transition is to
understand how a group of individuals becomes a new
kind of individual, possessing the properties of heritable variation in fitness at a new level of organization—
tantamount to evolvability of the new evolutionary individual. During evolutionary transitions, preexisting
individuals form groups, within which interactions
occur that affect the fitnesses of both the individuals
and the group. For example, under certain conditions,
bacteria associate to form a fruiting body, amoebae
associate to form a slug, solitary cells form a colonial
group, normally solitary wasps breed cooperatively,
birds associate to form a colony, and some mammals
form societies. In addition, about 2 billion years ago,
archaebacteria-like cells (destined to be the ancestors
of all eukaryotes) began alliances with other bacteria
to form the first eukaryotic like cell.
Such associations and groups may persist and reform with varying likelihood depending on properties
of the group and the component individuals. Initially,
group fitness is the average of the lower-level individual fitnesses, but as the evolutionary transition
proceeds, group fitness becomes decoupled from the
fitness of its lower-level components. Indeed, the
essence of an evolutionary transition in individuality
is that the lower-level individuals must “relinquish”
their “claim” to fitness, that is to flourish and multiply,
in favor of the new higher-level unit. This transfer of
fitness from lower to higher-levels occurs through the
evolution of cooperation and mediators of conflict that
restrict the opportunity for within-group change and
enhance the opportunity for between-group change.
Until, eventually, the group becomes a new evolutionary individual in the sense of being evolvable—
possessing heritable variation in fitness (at the new
level of organization) and being protected from the
ravages of within-group change by adaptations that
restrict the opportunity for defection (Michod, 1999).
2. A model for the evolution of an
early-segregated germ-line
2.1. Conflict mediation
Key in the conversion of a group of cooperators
to a new evolutionary individual is the evolution of
conflict mediators, genes that tweak development
R.E. Michod et al. / BioSystems 69 (2003) 95–114
and/or the ways in which the groups are organized
and by so doing tilt selection in favor of the group
and away from the level of the cell. Although the
underlying mechanisms may be diverse (germ-line,
apoptosis, self-policing, mutation rate, determinate
growth, reproductive mode and propagule size), conflict mediation serves to enhance the evolvability of
the higher-level unit by restricting the evolvability
of composite lower-level units. However, as we discover below in the green algal group Volvocales,
the way in which conflict mediation is accomplished
may be short-sited and end up interfering with
the long-term evolvability of the new multicellular
unit.
The steps involved in the origin of multicellular life
have been discussed by a variety of authors (see, for
example, Maynard Smith and Szathmáry, 1995). As
mentioned earlier, the initial step was likely the formation of cell-groups leading to some kind of primitive colonial life. These cell groups could have been
formed in several ways, such as through aggregation,
fragmentation, or single cell spore-like reproduction
(Michod and Roze, 2000, 2001; Roze and Michod,
2001). Single cell spore-like reproduction may have
been accomplished simply by the failure of the
daughter cells to separate following cell division. For
example, the multiple fission mode of reproduction in
volvocalean green algae discussed later is particularly
well suited to forming cell groups in this way. Along
with the formation of cell groups, there was likely the
evolution of cooperative functions which benefited
the group. Cooperative interactions are fundamental
to the emergence of new individuals, as only cooperation exports fitness from the lower to the new higher
level. However, the evolution of cooperation also sets
the stage for conflict. For multicellular organisms
to emerge out of cell groups, conflict mediation is
needed to regulate the levels of selection conflicts inherent in the initial structure of cell-groups (Michod,
1999). The developmental programs and organizational structures created through conflict mediation
are the first emergent functions at the higher level.
The evolution of a way of regulating conflict means
that the group is no longer a collection of lower-level
units, it is on its way to becoming a new higher-level
individual. For example, in the case of the evolution
of reproductive specialization through a germ and
somatic-line of cells (considered here), the group is
97
Fig. 1. Multi-level framework for the origin of multicellular organisms. The subscript j refers to the number of cooperating cells
in a propagule; j = 0, 1, 2,. . . , N, where N is the total number of cells in the offspring propagule group, assumed constant
for simplicity. The variable kj refers to the total number of cells
at the adult stage of propagules that start out with j cooperating
cells. The variable Wj is the fitness of group j, defined as the expected number of propagules produced by the group, assumed to
depend both on size of the adult group after development and its
functionality (or level of cooperation among its component cells)
represented by parameter β in the models later.
no longer divisible, because certain cells have special
functions.
2.2. Multi-level selection model
A multi-level selection approach begins by partitioning the total change in frequency of genotypes
into within and between-group components. Groups
are defined by a group property, usually the group frequency of a phenotype or genotype (or some other
property reflecting group composition). In our model,
we assume that offspring groups are composed of N
cells as in Fig. 1 (Michod, 1999; Michod and Roze,
1999, 2000). The level of kinship among cells in the
offspring groups is determined by N.
During development, cells proliferate and die (possibly at different rates depending on cell behaviour)
to create the adult cell group. Cell behaviour is assumed to be determined by a single haploid genetic
locus with two alleles, C (for cooperative cells) and
D for defecting cells. Deleterious mutation (from C
to D) may occur during cell division leading to the
loss of cooperative cell functions (such as the propensity to become motile in a Volvox colony considered
later) and a decrease in fitness of the adult group. The
98
R.E. Michod et al. / BioSystems 69 (2003) 95–114
adult group produces offspring groups of the next
generation. We have previously considered several
different modes of reproduction according to how
the offspring group is produced: single cell (or spore
reproduction), fragmentation and aggregation. In the
present model, we consider only single cell reproduction and assume that development starts from a single
cell (N = 1), as is the case in Volvox carteri and the
other members of the volvocine lineage (see Fig. 5
later).
The basic parameters of the model include development time, t, the within organism mutation rate from
C to D per cell division, µ, the effect of mutation on
the cell replication rate, b (b < 1 or b > 1 means uniformly deleterious or selfish mutations, respectively),
the deleterious effect of mutation on the cell group or
organism, β, and the propagule size, N. The parameter β measures the benefit of cooperation at the group
or organism level, and, hence, the deleterious effect of
mutation on group fitness. In addition, there is a parameter that tunes the relative effect of group size on
fitness.
To study how evolution may shape development
and the opportunity for selection at the two levels of
organization, we assume a second modifier locus with
two alleles M and m. The m allele encodes the ancestral state (no mediation) while the M allele changes
the parameters of development and/or selection at
the primary cooperate/defect locus. In this way, we
may understand the evolution of developmental programs that permit and enhance the evolvability of
the group. The conflict mediator M allele may affect
virtually any aspect of the model, such as propagule
size, N (studied elsewhere, Michod and Roze, 1999,
2000; Roze and Michod, 2001), adult size (whether
it is determinate or indeterminate, Michod and Li
unpublished results), self-policing, programmed cell
death, development time and whether there is an
early-segregated germ-line. To study the evolution of
self-policing, we assumed the modifier affects the parameters of selection at both levels, b and β, reducing
the temptation to defect at some cost to the group
(Michod, 1996; Michod and Roze, 1999). In the case
of the evolution of programmed cell death, we assumed the modifier directly decreases the replication
rate of mutant cells (Michod and Nedelcu, 2003). In
the present paper, we study how the M allele may
create an early-segregated germ-line.
2.3. Details of germ-line model
Using the two locus modifier model just outlined, we have studied the evolution of a sequestered
germ-line (Michod, 1996; Michod and Roze, 1997,
1999). We assume that the modifier M allele creates two separate lineages of cells within the group,
a germ- and somatic-line, which may each undergo
a different number of cell divisions and experience a
different mutation rate. In our previous work, we assumed that the germ-line was sequestered as a single
cell set aside during the first cell division. Most organisms depart from this ideal, including V. carteri, and
sequester cells later in development. For this reason,
we model the selective forces acting on the time of
sequestration, θ, and the number of cells sequestered,
ν. In our previous work, we assumed θ = ν = 1.
A critical assumption concerns the definition of fitness in groups with and without germ/soma differentiation. In groups without separate germ and somatic cells, all cells in the group perform reproductive
(germ) and vegetative (somatic) functions. We assume
fitness is proportional to group size, the number of
cells in the group, k, multiplied by a term representing the functionality of the group, (1 + βf ), where f
is the frequency of cooperative cells in the group and
β is the benefit of cooperation discussed previously.
This gives Eq. (1) for the fitness of groups without
germ/soma differentiation.
W = k(1 + βf)
(1)
In the groups with separate germ and somatic
cells, there must be fewer cells available for somatic
function, because some cells, ν, have been put aside
to produce the germ-line cells. Consequently, there
must be a cost to sequestering the ν cells; how should
we represent this cost during this early stage of
germ/soma evolution?
We consider the initial phases of germ/soma evolution in which the presence of a differentiated
germ-line is the only difference between the two
kinds of groups (groups made up of undifferentiated cells and groups made up of differentiated germ
and somatic cells). For example, we assume that the
resources available to the group remain essentially
unchanged by germ/soma differentiation, except as a
result of changes in the numbers of cooperative and
mutant cell types. The number of cell types change in
R.E. Michod et al. / BioSystems 69 (2003) 95–114
the differentiated group, because cooperative behaviors are expressed only by somatic cells. By virtue
of specializing in a single function, such as motility,
somatic cells may be able to perform the vegetative
functions better or for longer periods of time (after all
somatic cells do not have to take time to reproduce
the group). These benefits of specialization are clearly
important and make it easier to evolve germ/soma
differentiation, but we ignore these effects here; their
consequences seem obvious (in the sense that they
will make it easier to evolve germ/soma differentiation), and our interest lies elsewhere in the mediation
of conflict between levels of selection brought about
by the evolution of the germ-line.
To understand the cost of sequestering ν cells to
create the germ-line, we think of these ν cells in
terms of two descendent populations of cells, according to the conditions either of their new existence as
differentiated germs, or their prior existence as members of an undifferentiated cell group. In the case of
their prior existence, the ν cells would have given
rise to a descendent population of size, say, kν cells
undifferentiated with respect to reproductive or vegetative functions. In the case of their new existence
as differentiated germs, the germ sample (the ν cells)
replicates for perhaps a different period of time with
a different mutation rate, giving rise to a descendent
population of size, say, Kν germ cells. The cost of
the germ-line can be seen as stemming from either
the new Kν germ cells or the missing kν cells available for vegetative (somatic) functions.
Following the definition of fitness in undifferentiated groups (Eq. (1)), we assume that fitness of the
newly differentiated group is fecundity multiplied
again by a term representing the functionality of
the group (which has now changed because of the
germ-line cost). Fecundity is the number of gametes,
taken to be proportional to size of the germ-line Kν .
For concreteness, we think of the functionality as
the amount of resources accrued by the group for
reproduction and we let fitness equal the total number of gametes produced multiplied by the amount of
resources received by each gamete. We assume that
differentiation does not change the level of resources
available to the group, taken to be proportional to the
undifferentiated group size, k, except that there are
Kν new germ cells present, or kν somatic cells absent
(depending on how the germ-line cost is interpreted).
99
When the cost of the germ-line is based on the new
germ cells, fitness equals Kν (k − Kν )(1 + βfν )/Kν =
(k − Kν )(1 + βfν ), where fν is now the frequency of
cooperating cells in the soma (instead of in the total
group), giving Eq. (2).
Wν = (k − Kν )(1 + βfν )
(2)
Eqs. (1) and (2) were used previously in our study of
germ-line evolution when sequestration of a single cell
occurred during the first cell division (Michod, 1996,
1999; Michod and Roze, 1997, 1999, 2000). When the
cost of the germ-line is based on the missing somatic
cells, kν , fitness equals Kν (k − kν )(1 + βfν )/Kν =
(k − kν )(1 + βfν ) giving Eq. (3).
Wν = (k − kν )(1 + βfν )
(3)
In both Eqs. (2) and (3), the number of gametes
cancels and fitness depends only on the amount of
resource. In our model, producing many low quality
gametes is the same as producing a few high quality
ones. As shown later, depending upon how the cost of
a germ-line is paid (whether we use Eqs. (2) and (3)),
early sequestration of the germ-line may be advantageous or not.
As already mentioned, in many organisms with a
germ-line, the germ-line is sequestered not during the
first cell division but later in development. For example, in the green alga, V. carteri (considered in
more detail later), the precursors of the germ-line are
formed after five cell divisions, but the germ-line is
sequestered only after the ninth cell division. We extend our model to study the time of sequestration by
introducing two new parameters defined in Fig. 2, the
time when the segregation occurs (θ) and the number
of cells sequestrated to form the germ-line (ν). In our
previous model, the germ-line was differentiated from
the soma at the beginning of the development, from
Fig. 2. Time of germ–soma segregation. After θ divisions, ν cells
are sampled to form the germ-line, each cell in the germ-line
divides tG more times. The number of cell divisions in the soma
is t and δ = t − tG − θ is the difference in number of cell divisions
between the germ-line and the soma.
100
R.E. Michod et al. / BioSystems 69 (2003) 95–114
one single cell so that θ = 1, ν = 1 and tG was a free
parameter. Among the sample of cells sequestrated to
form the germ-line at time θ, there will be some C
cells and some D cells, the number of C cells being
given by a probability distribution. The technical details underlying this distribution and its implications
for the model are given in the Appendix A.
Using two locus modifier techniques outlined earlier
and discussed further in Appendix A, we have studied the evolution of germ-line modifiers that sequester
ν cells after θ cell divisions for the two kinds of fitness functions (two kinds of germ-line cost) given in
Eqs. (2) and (3). In both formulations, fitness depends
upon the resources available to the germ-line, however
in Eq. (2), the cost of the germ-line depends upon the
number of gametes produced by the germ-line, while
in Eq. (3), the cost depends on the missing cells that
are no longer available for vegetative function.
3. Results
Results using fitness Eq. (2) are given in Fig. 3 for
the following parameter values µ = µG = 0.003,
β = 3, b = 1.05, t = 30 and no sex or recombination
(r = 0). These parameter values have been explored
Fig. 3. Evolution of germ-line segregation using fitness Eq. (2). In Eq. (2), we represent the cost of the germ-line by subtracting off
the number of resulting germ cells (after the ν cells replicate for tG times), k − Kν . The germ-line modifier is characterized by three
parameters: the time of sequestration, θ, the number of divisions after sequestration, tG , and the number of cells sampled, ν. Germ-line
evolves for parameter values below the surfaces in all panels. The total number of replications in the germ is tG + θ. The parameters for
the three slices in the two dimensional plots are given from top to bottom. For example, in panel (D), the top solid curve is for θ = 5
and the bottom dotted curve is for θ = 20. Propagule size N = 1. Methods used to construct the graphs are given in the Appendix A.
R.E. Michod et al. / BioSystems 69 (2003) 95–114
and discussed in our previous work and they are used
here for comparison purposes. Please consult our previous papers for discussion of the rationale for using these parameter values (for example, Michod and
Roze, 2000). The germ-line modifier evolves for parameter values below the surfaces plotted in Fig. 3. In
our previous work, the germ-line modifier was characterized by a single parameter, δ, the difference in
number of cell divisions between the germ-line and the
somatic-line. However, the germ-line modifier is now
characterized by three parameters: the time of sequestration, θ, the number of divisions after sequestration,
tG , and the number of cells sampled, ν. The difference
in number of cell divisions between the germ-line and
the soma, δ = t − tG − θ, now depends on two of the
germ-line parameters in addition to the soma parame-
101
ter, t. The number of cell divisions after segregation of
the germ-line, tG , is important in that it (in conjunction with the number of cells sampled, ν) determines
the size of the germ-line and ultimately the number of
gametes, that is this parameter determines fecundity
(see Fig. 8 later).
The results in Fig. 3 are straight forward; it is easier for a germ-line to evolve (larger germ-lines may
be produced) the earlier the germ-line is sequestered
(Fig. 3, especially panel (B)), the lower the number
of times it divides, and the fewer number of cells that
are sampled. This is because there are only advantages
to early segregation and low replication (in terms of
a lower effective deleterious mutation rate resulting
from the fewer number of cell divisions), and the cost
of the germ-line is smaller the fewer cells that are
Fig. 4. Evolution of germ-line segregation using fitness Eq. (3). In Eq. (3), we represent the cost of the germ-line by subtracting off the
missing somatic cells, k − kν . Germ-line evolves for parameter values below the surfaces in all panels. See legend to Fig. 3. Propagule
size N = 1. Methods used to construct the graphs are given in the Appendix A.
102
R.E. Michod et al. / BioSystems 69 (2003) 95–114
sampled. Recall, that in our formulation, fitness does
not depend on the size of the germ-line directly (only
through resource availability), since the number of gametes cancels out in the formulation of fitness. Organisms following this model (using Eq. (2)) should form
a germ-line by sequestering a single non-dividing cell
during the first cell division.
What about when the resources available to the
germ-line depend on the number of cells missing unavailable for vegetative (somatic) function? The missing cells are those that would have been formed by the
ν cells sequestered to form the germ cells. In Fig. 4,
we show the results for the same set of parameter values as used in Fig. 3, except using Eq. (3) instead of
Eq. (2). In this case, there is a cost to the germ-line in
terms of missing somatic cells, and, so, there is an intermediate optimum time for sequestration, θ (Fig. 4,
especially panels (B) and (C)). Smaller θ is better in
terms of coping with the threat of deleterious mutation, however, there is a greater penalty to pay in terms
of missing cells unavailable for somatic function.
4. Volvocalean green algae as a study case
4.1. The volvocalean green algal group
We use the volvocalean green algal group to test
our model for the evolution of an early segregated
germ-line. The volvocalean green algae comprise
both unicellular (Chlamydomonas-like) algae as well
as colonial forms in different stages of organizational
and developmental complexity (Fig. 5). Interestingly,
both multicellularity and germ/soma separation have
evolved multiple times in this group. The different
levels of organizational and developmental complexity are thought to be alternative stable states (Larson
et al., 1992). Nevertheless, despite their multiple and
independent acquisition of the multicellular state and
germ/some separation, none of these multicellular
lineages attained high levels of complexity, as did the
green algal ancestors of land plants, the charophytes.
We believe that understanding this limited spurt of diversification in complexity in this lineage will provide
insight into how a transition to multicellularity may
affect the evolvability of the new multicellular unit
(Nedelcu and Michod, 2003). We suggest here that,
although a germ-line acts to mediate conflict between
Fig. 5. The volvocine lineage. A subset of colonial volvocalean
green algae that show a progressive increase in cell number, volume of extracellular matrix per cell, division of labor between
somatic and reproductive cells, and proportion of vegetative cells.
A: Chlamydomonas reinhardtii; B: Gonium pectorale; C: Pandorina morum; D: Eudorina elegans; E: Pleodorina californica; F:
V. carteri. Where two cell types are present, the smaller cells are
the vegetative/somatic cells, whereas the larger cells are the reproductive cells (gonidia). Images kindly provided by David L. Kirk.
the group and cell levels, and is thereby expected to
contribute to the evolvability of the new multicellular
group, the way in which the germ and somatic-lines
are created may nevertheless interfere with the longer
term evolvability of the lineage.
The basic morphological and developmental traits
of the volvocalean green algae appear to result from
the interaction of conflicting structural constraints (imposed by two very conserved anatomic traits) as well
as two strong selective pressures. The first structural
constraint is the so-called “flagellation constraint”
(Koufopanou, 1994) which leads to a trade-off between reproduction and motility. This constraint has
a different structural basis than the one invoked in the
origin of metazoans (Margulis, 1981; Buss, 1987),
although the end result is similar. In most green flagellates, during cell division the flagellar basal bodies
remain attached to the plasma membrane and flagella,
and can act as centrioles (which is not the case in
animal cells); however, in volvocalean algae, due to a
coherent rigid cell wall the position of flagella is fixed
and thus, the basal bodies cannot move laterally and
take the position expected for centrioles during cell
division while remaining attached to the flagella (as
they do in other green flagellates). Consequently, cell
division and flagellar motility can take place simultaneously only for as long as flagella can beat without
R.E. Michod et al. / BioSystems 69 (2003) 95–114
having the basal bodies attached (i.e. up to five cell
divisions). The second constraint comes from the
unique way of cell division in volvocalean green algae,
namely, the multiple fission type of division termed
palintomy: cells do not double in size and then undergo binary fission; rather, cells grow about 2n -fold
in volume, and then rapidly undergo a synchronous
series of n divisions (under the mother cell wall).
Because clusters, rather than individual cells, are produced in this way, this type of division is suggested
to have been an important precondition facilitating
the formation of volvocacean colonies (Kirk, 1998).
The two selective pressures that are thought to have
favored an increase in complexity in volvocalean algae are the advantages of a large size and the need
for motility (Bell, 1985; Kirk, 1998). Larger size is
thought to be advantageous by allowing colonies to
escape predators, move faster, maintain better homoeostasis or better exploit eutrophic conditions,
while motility allows the colonies better access to the
euphotic/photosynthetic zone. Interestingly, as discussed next, given the ancestral constraints just mentioned, namely the flagellar constraint and the multiple
fission type of cell division, it is difficult to achieve
the two selective advantages simultaneously. Larger
size via higher number of cells requires increased total time for cell division and, when coupled with the
flagellation constraint, this means decreased motility.
103
4.2. Germ/soma separation in Volvox
This negative impact of the flagellation and palintomy constraints was likely overcome by a division
of labor (cellular differentiation) in the colony: some
cells became specialized for motility, while others
were assigned to reproductive functions. The proportion of cells that maintain motility and become sterile
is directly correlated with the total number of cells
in a colony: from none in Chlamydomonas, Gonium,
and Eudorina, to up to one-half in Pleodorina and
over 99% in Volvox (Larson et al., 1992). In the latter,
the division of labor is complete: only the gonidia
(the reproductive cells) undergo cleavage to form
new colonies; the somatic cells are sterile, terminally
differentiated, and are thought to be genetically programmed to undergo cellular senescence and death
once the progeny was released from the parental
colony (Pommerville and Kochert, 1981, 1982).
Volvox represents the most complex multicellular
form in the volvocine lineage (Fig. 5). Germ/soma
separation evolved at least twice, but possibly several
times in Volvox (Kirk, 1999) and is realized differently
among the at least 18 recognized species of Volvox
(Desnitski, 1995). In some Volvox species, including
V. carteri, the two types of cells are set apart by asymmetric divisions early in the embryonic development.
Moreover, in V. carteri a difference in size, not a difference in cytoplasmic quality, determines which pathway of differentiation a cell will follow (Kirk et al.,
1993). It should be mentioned that our discussion later
is relevant only to this particular species of Volvox.
4.2.1. Overview
As the volvocalean colonies increase in size and
number of cells, the number of cell divisions as well as
the size of the mature reproductive cell also increase.
Due to the flagellation and palintomic constraints discussed above, the motility of the colony during the
reproductive phase is negatively impacted for longer
periods of time than are acceptable in terms of the
need to access the photosynthetic zone. For instance,
in a V. carteri mutant with only 256 cells and no
germ/soma separation (discussed later), the flagellar
motility of the colony is negatively affected for as
much 70% of the life cycle (i.e. 49–50 h in a 72-h life
cycle); its decreased motility might be responsible for
this mutant not being found in nature. In short, the
need for the colony to increase in size detracts from
the motility of the colony.
4.2.2. A comparative view
Many multicellular organisms have a germ-line that
is segregated early in the development (for a list see
for example Table 1.1 in Buss, 1987). Among these,
the nematode worm Caenorhabditis elegans and the
green alga V. carteri have been characterized in detail
in terms of their development. Interestingly, although
the number of somatic cells is small in both organisms (i.e. 959 cells versus 2000–4000, respectively),
there are major differences between the two species
with respect to the way in which germ-line is produced
(Fig. 6) and the number of cell types (more than 20
in the former and only 2 in the latter). What might be
the evolutionary reasons for such differences? Later,
we describe features of development in the two lineages, contrast them, and point out the peculiarities of
germ/soma separation in V. carteri.
104
R.E. Michod et al. / BioSystems 69 (2003) 95–114
Fig. 6. Two examples of early germ-line segregation. (A) Germ-line segregation in the nematode worm C. elegans; gray, black, and white
ellipses indicate germ-line blastomeres, primordial germ cells, and somatic blastomeres, respectively. P0 denotes the zygote; P1, P2, P3,
are the germ-line blastomeres; P4 is the germ-line founder cell; Z2 and Z3 are primordial germ cells. Numerals indicate successive cell
divisions. (B) Germ-line segregation in the green alga V. carteri; gray and white large ellipses denote totipotent and somatic blastomeres,
respectively; gray, black, and white small ellipses indicate germ-line blastomeres, gonidia, and somatic initials/cells, respectively. Details for
the division pathways are shown only for one germ-line founder cell and one somatic blastomere. Numerals mark successive cell divisions.
The first four cell divisions during the C. elegans
embryonic development are asymmetric (e.g. Seydoux
and Schedl, 2001). The zygote, P0, as well as the
smaller cells resulting from the first three asymmetric divisions (i.e. P1–P3 in Fig. 6A) act as stem cells:
with every asymmetric cell division they produce another totipotent stem cell (they renew themselves) and
a somatic blastomere. These stem cells, or germ-line
blastomeres, are the precursors of the germ-line. The
last of these cells, P4, represents the founder cell (or
progenitor) of the germ-line. Ultimately, the P4 cell
divides symmetrically to produce two primordial germ
cells (PGCs), Z2 and Z3, that migrate to the gonad
and arrest mitosis until the larva hatches; later on, Z2
and Z3 proliferate and undergo meiosis to produce
the germ cells. What specifies the fate of the cells
following the four asymmetric divisions is a set of
germ-line granules, or P granules, which are present
in the zygote and are selectively distributed only to the
germ-line precursors and ultimately to the two PGCs
(e.g. Seydoux and Schedl, 2001). The larger cells resulting from the first four asymmetric divisions act as
somatic blastomeres or founder cells, and proliferate
to form various cell types, such as neurons, epithelial,
muscle, gland cells, etc.; some of the divisions responsible for the 959 somatic cells in the adult take place
after the larvae hatch.
In contrast to C. elegans, the first divisions in V.
carteri are symmetrical (Fig. 6B). Up to the 16-cell
stage, all the blastomeres are totipotent (if removed
they can create a new colony); after the fifth division,
the 16 cells situated at the anterior pole of the embryo
will take a distinct path relative to the 16 cells at the
posterior pole. The first difference is that the sixth
cell division is asymmetric in the 16 anterior cells
but symmetric in the other 16 cells. What determines
which cells will undergo an asymmetric division is
not well understood; physical cues (i.e. physical constraints) associated with the number of cytoplasmic
bridges and the position of cells in the embryo are
thought to be involved in this process (Kirk, 1994).
By having the first blastomeres functionally equivalent (i.e. totipotent) and by relying on spatial cues for
initiating the pathway leading to the differentiation
of the germ-line, V. carteri differs from C. elegans
but resembles mammals (e.g. Anderson et al., 2001).
R.E. Michod et al. / BioSystems 69 (2003) 95–114
However, the further steps in the segregation of the
germ-line are rather unique in V. carteri.
The 16 cells predetermined to produce the germ-line
are totipotent, act as stem cells and undergo a series
of three–four successive asymmetric cell divisions, in
a manner similar to that of the C. elegans germ-line
precursors (germ-line blastomeres P0–P3) (Fig. 6).
Nevertheless, the similarities stop here. In contrast to
C. elegans, in V. carteri, the larger cell resulting from
the asymmetric division is the one to act as a germ-line
blastomere. Furthermore, the asymmetric division is
not accompanied by a differential segregation of a
specific cytoplasmic component; what triggers the somatic fate of the smaller cells in V. carteri is not the
lack of distinct cytoplasmic components, as it is in C.
elegans, Drosophila, and Xenopus (e.g. Kloc et al.,
2001; Mahowald, 2001; Seydoux and Schedl, 2001),
but rather strictly a difference in cell size (Kirk et al.,
1993). In addition, the large cells that result from
the ninth division (and which are analogous to the
C. elegans’s progenitor of the germ-line, P4) do not
divide to produce mitotically active primordial germ
cells, PGCs. Rather, these cells arrest mitosis (while
the somatic blastomeres continue to divide for another
two–three cycles) and differentiate without further
divisions into single non-flagellated germ cells, gonidia. Nevertheless, although each germ-line founder
cell produces only one germ cell (gonidium), it is
interesting that there are up to 16 such cells in each
spheroid, as if there were 16 independent germ-lines
each producing a single germ cell. On the other hand,
the smaller cells produced though asymmetric cell
divisions in the anterior pole as well as the cells produced by the 16 posterior blastomeres stop dividing
while still in the embryo, and all differentiate into the
same type of cell, i.e. flagellated somatic cells.
5. Germ/soma separation and evolvability during
the transition to multicellularity
We have discussed elsewhere how the transition
from unicellular to multicellular life requires the decoupling of basic life-properties at the lower level and
their re-coupling and re-organization in new ways at
the higher level (Nedelcu and Michod, 2003). Moreover, we argued that some of the differences among the
extant species, including differences in evolvability,
105
can be explained by the way in which the decoupling
and re-coupling of these properties has been achieved
during the transition to multicellularity.
Two of the complex life-traits that become reorganized during the unicellular–multicellular transition are immortality and totipotency. By “immortality”
we mean the continued capacity for cell division and
by “totipotency” we mean the capacity of a cell to
create a new organism. In unicellular organisms, these
traits are necessarily coupled at the cell level in the
sense that they are both fully expressed in all cells.
In multicellular organisms, however, the two traits
are re-organized within and between cell lineages.
In species with defined germ and somatic-lines, the
two traits are fully expressed only in the germ-line,
whereas somatic cells obviously do not express
totipotency and express continued replicative ability,
that is immortality, to varying degrees. In V. carteri,
immortality and totipotency are fully restricted to the
germ-line and somatic-lineages have no mitotic or
replicative potential (Nedelcu and Michod, 2003). The
un-coupling of immortality and totipotency proved
not possible in V. carteri: these traits are express either together and fully (i.e. in the gonidia) or not at all
(i.e. in the somatic cells). Immortality and totipotency
are thus still tightly linked in V. carteri, as they are
in their unicellular ancestors. In support of this view
is the fact that “cancer-like” mutant somatic cells, in
which immortality but not totipotency is re-gained,
are missing in V. carteri. There are, however mutant
forms of V. carteri (discussed later) in which somatic
cells re-gain both immortality and totipotency, but in
neither of these mutants are the two traits expressed
partially or differentially (e.g. limited mitotic capacity
or multipotency).
To ensure the emergence of individuality at the
higher level and the reproduction of the multicellular individual (i.e. the heritability of the group-level
traits), immortality and cell division have to be
de-coupled from the reproduction of the lower-level
cells (previously unicellular individuals) and be coopted for the reproduction of the group (the higherlevel multicellular unit). Furthermore, in lineages with
a separation between germ and soma, in somatic cell
lineages, cell division is not associated with reproduction of the individual but rather became co-opted
for growth of the multicellular individual (Nedelcu
and Michod, 2003).
106
R.E. Michod et al. / BioSystems 69 (2003) 95–114
V. carteri avoided the risks and potential conflicts
of cell division in somatic cells by blocking it altogether and in this way achieved individuality at the
cell-group level. In addition cell division was not
co-opted for the post-embryonic growth of the multicellular individual. We believe that this solution to
the mediation of conflicts at the lower level affected
the long-term evolutionary adaptability of the lineage. Later, we investigate the possible reasons for
undertaking this peculiar pathway. Our fundamental
point is that, although conflict mediation is pivotal
to the emergence of individuality at the higher-level,
the way in which the mediation is achieved during
the transition in individuality can interfere with the
long-term evolvability of the lineage.
6. The evolution of a germ-line in Volvox carteri
6.1. The question
In many lineages, including C. elegans for instance,
the germ cells are the descendants of a single germ
cell founder (e.g. P4 in C. elegans; Fig. 6A). In contrast, as we discussed earlier, in V. carteri, the 16 germ
cells are formed from 16 independent germ-line blastomeres that stop dividing and differentiate directly
into gonidia (Fig. 6B). Theoretically, the same 16 go-
nidia could be produced from a single founder cell (as
is the case in C. elegans) by dividing a total of four
times (1 × 24 = 16); this raises the question as to why
in V. carteri the germ cells differentiate from 16 independent germ-line founder cells instead of only one
germ founder cell.
To address this issue, we envisioned for V. carteri a developmental pattern similar to that of C. elegans, and asked what the consequences would be.
Surprisingly, such a pathway would be identical to
the current one in terms of the numbers of gonidia
and somatic cells (Fig. 7). We envisioned that after the fifth cell division (which is the time where
the 16 germ-line blastomeres are determined in V.
carteri), only one cell is sequestered, and would divide symmetrically a total of four times to produce
16 gonidia; the other 31 cells undergo a path identical to that of the 16 posterior cells in V. carteri,
that is they divide symmetrically seven times to produce somatic cells (Fig. 7). Interestingly, by having
only one cell acting as a germ-line founder and 31
(instead of 16) cells acting as somatic blastomeres,
the same total of 3968 somatic cells (i.e. 31 × 27 =
3968) and 16 gonidia (i.e. 1 × 24 = 16) would be
produced. Furthermore, under this scenario, asymmetric divisions are not required (see discussion later).
Then, why was the former alternative favored over
the latter, especially when it requires the evolution
Fig. 7. Two alternative scenarios for germ-line segregation in V. carteri. (A). The alternative that has been selected for in V. carteri:
germ cells differentiate from 16 independent germ-line founder cells. (B). Our theoretical alternative: germ-line cells derive from a single
germ-line founder cell. Symbols are as in Fig. 6.
R.E. Michod et al. / BioSystems 69 (2003) 95–114
of a new trait, namely, the asymmetric type of cell
division?
6.2. Applying the model
Can the model on the evolution of an early segregated germ-line presented earlier help address this
question? We use our model to ask under which conditions would one or the other alternative diagrammed in
Fig. 7 be selected for, so as to understand the “choice”
that V. carteri has made.
6.2.1. Assumptions
In trying to understand the various selective factors,
we assume that 12 cell divisions is the highest number
of cell divisions possible under the multiple-fission
palintomy type of cell division. In reality, this is
true for V. carteri. As discussed earlier, with the
multiple-fission palintomy, to reproduce a colony with
2n cells, a single reproductive cell must first grow 2n
before it divides n times. Clearly, a single cell cannot
grow indefinitely in size (especially if it is a spherical
shape) without affecting its metabolic abilities (due
to changes of the surface/volume ratio). It appears
that a 500–1000-fold increase in size is the most a
gonidia can stand without affecting its metabolism,
and 212 cells is the highest number of viable cells
that can result from such a gonidia. We do not have a
detailed argument for this assumption, however there
are no palintomic Volvox species in which n is higher
than 12. Nevertheless, in Volvox species that have
“escaped” the palintomic constraint, n can be higher
than 12. As a consequence, gonidia do not have to
grow up to 2n and then divide n times, rather they only
double in size, divide, grow between divisions, divide
again and so on. Interestingly, in the non-palintomic
species, gonidia differentiate late in development,
suggesting a correlation between palintomy and
early-segregation of germ (i.e., the fast cell division
cycles associated with palintomy may cause higher
mutation rates).
6.2.2. Cost of the germ-line
We interpret the cost of the germ-line in terms of the
missing somatic cells that are unavailable for vegetative functions (as expressed in Eq. (3)). In V. carteri,
such a cost is likely very important, due to the selective pressures mentioned previously, namely, the need
107
for daily vertical migrations in the water column and
a large body-size. The presence of a non-flagellated
germ-line affects the fitness of the group by interfering
with the mobility of the group through decreasing the
number of cells that participate in motility (i.e. by affecting the ratio of flagellated to non-flagellated cells).
In addition, an early segregated and arrested germ-line
decreases the overall number of cells in the group and
thus the body-size is affected.
The time of sequestration as well as the number of
cell divisions following the sequestration event influence the cost of the germ-line both in terms of the
total number of cells missing in the group as well as
the missing soma, that is missing flagellated cells. We
illustrate the trade-offs involved with three examples.
First, if only one germ-line founder cell is sequestered
after the first embryonic cell division (as in our earlier model) and continues to divide for the same 12
times as the somatic-lineages, the total number of cells
would not be affected (i.e. 4096), but the flagellated
to non-flagellated cell ratio would be 1:1 and thus the
motility of the group would be greatly affected. In this
case, the cost of the germ-line is only in terms of the
number of flagellated cells missing (and not overall
colony size), but the cost is very high (i.e. 2048 missing motile cells). In addition, the germ cells would
likely accumulate the same number of mutations as the
somatic cells assuming mutation is dependent on the
number of DNA replications. Second, if the founder
cell segregated after the first cell division and underwent only four cell divisions to produce 16 gonidia
(and then arrest mitosis while the somatic-lineages
continue dividing for a total of 12 times) the group
would be composed of only 2048 somatic cells and
16 gonidia. In this case, the ratio increases in the favor of the motile cells (2048:16), but the colony size
is affected (2064 versus 4096 cells). The cost of having a germ-line segregating very early and arresting
mitosis after four cell divisions is very high in this
example, both in terms of total number of cells (i.e.
2032 missing cells) as well flagellated cells (i.e. 2048
cells). However, germ cells would have undergone a
fewer number of cell divisions than the somatic cells
and would have fewer mutations as a result. Third,
at the other extreme, if the founder cell is put aside
only after the eighth division and undergoes four additional cell divisions to produce the 16 gonidia, the cost
would only be in terms of the flagellated cells, which
108
R.E. Michod et al. / BioSystems 69 (2003) 95–114
is very low (i.e. 16 cells). Nevertheless, the germ cells
would be dividing for the same number of times as the
somatic cells.
A compromise may be reached between group
size and percentage of flagellated cells by having the
germ-line segregate at an intermediate stage during
development and arresting mitosis earlier than in the
somatic-lineages, such that both the cost in terms of
missing cells and the mutation levels are lower. If
only one founder-germ cell segregates after the fifth
division (Fig. 7B) and undergoes a number of additional four divisions to produce the 16 gonidia, the
cost of germ-line in terms of total number of cells
is 112 cells (4096 − 3984 = 112) and in terms of
flagellated somatic cells is 128 cells (4096 − 3968 =
128). In addition, there is a 25% decrease in the
number of cell divisions in the germ cells relative
to somatic cells (9 versus 12), which might result
in a lower mutation level in the germ-line. However, as discussed later, the same low cost can be
achieved through either the scenario just discussed
or the pathway that V. carteri is actually following
(Fig. 7A); why the latter was selected for is discussed
next.
6.2.3. Parameters and values
We used two sets of values for the model’s parameters: first, the values used by V. carteri, and second,
the values that correspond to the theoretical alternative we envisioned in Fig. 7B. In both cases, the
somatic cells divide for up to 12 times, thus t = 12
in our model. Likewise, the cost of the germ-line
is the same in both cases following that assumed in
Eq. (3) earlier; if all cells remained undifferentiated
and replicated 12 times, there would be 4096 cells in
the group, so there are 128 (i.e. 4096 – 3968) cells
missing in the soma. We assume asexual reproduction
(r = 0), a significant benefit of cooperation (coming
from motility) β = 10, and that the mutation rates per
cell division are equal in the germ and soma, µ = µG .
In addition, for each set of conditions, we used two
values for the mutation rate, µ = 0.003 and an order
of magnitude higher µ = 0.03 (and, as already mentioned, the number of missing somatic cells is the
same in both cases). The difference between the two
situations concerns the sampling of mutations and the
resulting opportunities for conflict between levels of
selection.
The parameters that differ in value between the
two cases are the time of sequestration (θ), the number of germ-line cell divisions after sequestration
(tG ), and the number of cells sampled (i.e. germ-line
founders, ν). In the first case (Fig. 7A), there are
16 germ-line founder cells (ν = 16) that segregate
from the somatic-lineage after the ninth cell division
(θ = 9) and arrest mitosis at that very point (tG = 0).
In contrast, in the second case (Fig. 7B), there is only
one germ-line founder cell (ν = 1) that segregates
earlier (θ = 5) and divides for an additional four
times (tG = 4). In both cases, the total number of cell
divisions that gonidia undergo is nine, while the additional number of cells divisions in the somatic-lineage
is three (δ = 3).
6.2.4. Predictions
Fig. 8 depicts the results of our model with both
sets of values (panels A and C versus panels B and
D) and under two mutation rates (panels A and B
versus panels C and D). Interestingly, the model predicts that a germ-line is easier to evolve (the germ-line
evolves for a greater range of parameter values) under the second set of parameter values, namely, those
corresponding to a germ-line descending from a single germ-cell founder, the route not taken in V. carteri.
In both cases, the model predicts that the germ-line
evolves easier when mutation is a threat, either because mutations are selfish (b > 1) or frequent (a high
mutation rate). When mutations are frequent, cooperation is harder to maintain, and so the within-group
advantage for the mutant cells (represented by parameter b) has to be lower. Under a lower mutation rate,
a germ-line can evolve only when the within-group
advantage of mutant cells is rather high, approaching
1.2 in the first case, or 1.1 in the second case. Also,
in the first case (Fig. 8 panel C), a germ-line evolves
only if the number of cell divisions after segregation
is very small, approximately tG must be less than 2.
These conclusions are mirrored in Fig. 4. Panel C of
Fig. 4 shows that tG (fecundity) can be higher with
ν = 1 and θ = 5 than with ν = 16 and θ = 9, and
this explains the results of Fig. 8. It is always good to
decrease ν (the initial number of germ cells), and it is
good to decrease θ up to a certain point, after which
there are not enough somatic cells (this is why we see
a maximum on Fig. 4C). So we can not say a priori
if ν = 1 and θ = 5 is better than ν = 16 and θ = 9
R.E. Michod et al. / BioSystems 69 (2003) 95–114
109
Fig. 8. Evolution of germ-line segregation in V. carteri. The main contrast in the figure is between a germ-line developed from 16 founder
cells sequestered at cell division 9 (panels (A) and (C)) and a germ-line from a single founder cell sequestered at cell division 5 (panels
(B) and (D)). The ordinate is tG (as it is in Figs. 3 and 4) except expressed in terms of fecundity, that is the number of resulting gonidial
germ cells. For example, if a single cell is sequestered (ν = 1) to divide five times (tG = 5), fecundity is 16 gonidia. Likewise, if 16
cells are sequestered (ν = 16) to divide no times (tG = 0), fecundity is 16 gonidia. The mutation rate is an order of magnitude higher
in the bottom panels than in the top panels. In the region labeled “no cooperation” group living cannot be maintained. In the two other
regions, cooperative groups are maintained with or without a germ-line. All panels assume fitness Eq. (3) (cost of the germ-line results
from missing somatic cells), equal mutation rate in germ and soma (µ = µG ), no sex (r = 0), benefit of cooperation β = 10, number
of cell divisions in soma t = 12. Recall tG is the number of cell divisions in the germ-line after sequestration and b is the within-group
advantage of mutants. Noting δ = t − tG , compare Figs. 6–1 (Michod, 1999), Fig. 1 (Michod and Roze, 1997), Fig. 6 (Michod and Roze,
1999). Methods used to construct the graphs are given in the Appendix A.
(because of the non linear effect of θ), but Figs. 4 and 8
show that it should be better. However, the developmental biology of Volvox carteri does not agree with
this prediction of the model.
In summary, the model predicts that in V. carteri a
germ-line is easier to evolve if mutant cells are selfish
(but not so selfish that cooperation cannot be maintained), or if the mutation rate is rather high (Fig. 8).
Are any of these conditions met in V. carteri? Moreover, the model suggests that the scenario based on
a single germ-cell founder (Fig. 7B) should be easier
to evolve than the one that has actually evolved in V.
carteri (Fig. 7A). Then, why was the germ-line in V.
carteri not formed under the conditions (i.e. a single
germ-cell founder) that the model predicts are most
conducive for the evolution of a germ-line? Below we
discuss these issues.
7. Discussion
7.1. Selfish mutants
According to the model (Fig. 8), a germ-line is
predicted to evolve only when the advantage of
within-group mutations is rather high. This is consistent with the fact in V. carteri many mutations that
affect the somatic cells are selfish. In the somatic regenerator mutants, or Reg mutants, the somatic cells
start out as small flagellated cells (wild type-like) and
then enlarge, loose flagella and re-differentiate into
gonidia. A number of 39 mutants in four phenotypic
classes have been investigated, and all had mutations at the same locus, regA (Huskey and Griffin,
1979). The gene affected in these mutants has been
shown to encode for an active repressor (Kirk et al.,
110
R.E. Michod et al. / BioSystems 69 (2003) 95–114
1999) that targets a number of at least 13 nuclear
genes whose products are required for chloroplast
biogenesis (Choi et al., 1996; Meissner et al., 1999b).
This finding suggests that the mechanism for the
establishment of a stable germ-soma separation in
V. carteri is based on preventing the somatic cells
from growing (by repressing chloroplast biogenesis
in these cells (Meissner et al., 1999a)) enough to
trigger cell division. In another class of mutants, the
Gls/Reg mutants (Huskey and Griffin, 1979), all the
cells (though fewer than in the wild-type, i.e. no more
than 128 or 256) act first as somatic cells and then
re-differentiate into reproductive cells; these mutants
resemble volvocacean species with no germ/soma
separation, such as Eudorina. The Gls mutation has
been mapped to a gene, glsA, that encodes a protein
required for the asymmetric divisions responsible for
the segregation of germ-line and somatic blastomeres
(Fig. 5A) (Miller and Kirk, 1999). Consequently, all
cells are equal both in size and potential for differentiation, and undergo the ancestral Chlamydomonas-like
pathway of acting first as vegetative and then as reproductive cells (Tam and Kirk, 1991); it should be
mentioned that this mutation is only recovered on a
regA− background such that the growth of somatic
cells and thus their differentiation into reproductive
cells is not suppressed.
The expression of RegA and Gls/Reg mutations has
a profound effect on the higher-level unit, that is the
colony. While the somatic cells re-differentiate into
gonidia, the spheroid is unable to maintain its motility (and thus its position in the euphotic zone) for
more than the duration of the first five cell divisions;
however, the total number of cell divisions required
to reach the final number of cells in the embryo is
seven or eight, and eleven or twelve in the Gls/Reg
and Reg mutants, respectively. In conditions where
motility and access to light are strong selective pressures, the higher-level is negatively affected by the
occurrence of these mutant cells at the lower-level,
which are thus acting as “selfish” mutant cells. To argue for the negative effects at the higher-level of these
types of selfish mutations in the environments where
wild-type forms of Volvox are usually present is the
fact that neither of the mutant forms are found as established populations in nature, although the Reg mutants occur spontaneously at a rather high rate (Kurn
et al., 1978). Interestingly, however, when access to
light and the need for motility are not representing selective pressures (i.e. in lab settings or possibly shallow waters), the fitness of these mutant forms proves
higher than the fitness of the wild-type (Koufopanou
and Bell (1991) and our unpublished data).
7.2. Mutation rates
In our model, we used a value for the genome wide
mutation rate reported for unicellular organisms such
as yeast, bacteria, viruses (Drake, 1991), yet the model
studies only a single locus assumed to represent the
genome wide functions involved in cellular functions.
Our use of a genome wide rate in a single locus model
is clearly a leap, but it gives an idea of what we hope
the model represents. We have considered more realistic mutation models elsewhere (Michod and Roze,
2000; Roze and Michod, 2001). Concerning the higher
mutation rate assumed in panels C and D of Fig. 8, it
is worth noting that volvocalean green algae seem to
feature levels of nucleotide substitution (as suggested
by the differences in branch length observed in phylogenetic trees based on nuclear rRNA sequences (e.g.
Friedl, 1997; Nakayama et al., 1998) that are higher
than those in other green algae as well as in their close
relatives, the land plants, which, incidentally, do not
have an early-defined germ-line.
7.3. Sixteen versus one germ-line founder
Although our model predicts that in both cases a
germ-line is more likely to evolve if the mutation
rate and the threat of selfish mutations are higher,
these conditions are more relaxed under the scenario
in which gonidia descend from a single germ-line cell
founder (Fig. 8B), instead of 16 different founder cells
as is the case in V. carteri. Why is this scenario not the
one that was selected for in V. carteri? Interestingly,
although the end result, as far as the composition (i.e.
number and type of cells) of the colony, is the same
under both scenarios, two aspects are different and not
represented in the model.
One aspect not represented in the model is asymmetric cell division. The second scenario (Fig. 7B)
does not use or require asymmetric divisions. However, the scenario that has been selected for in V. carteri
involves asymmetric divisions; so, why did asymmetric division evolve in V. carteri? Asymmetric division
R.E. Michod et al. / BioSystems 69 (2003) 95–114
is considered a derived trait (Desnitski, 1992) in the
volvocalean green algal group. Asymmetric divisions
do not occur in the less complex Pleodorina as well as
many species of Volvox; in these cases, all gonidia are
developed from cells that were initially indistinguishable from somatic cells, but then undergo enlargement
and differentiation. In V. carteri, asymmetric divisions
and the segregation of the germ-line are not associated
with the differential segregation of cytoplasmic components (Kirk et al., 1993). Rather, it has been shown
that the cells that remain above a threshold (i.e. 8 m)
differentiate into germ-cells, whereas the cells that fall
below the threshold will take the somatic path (Kirk
et al., 1993), regardless of the type of cell division. To
argue for this is also the fact that the formation of the
2048 somatic cells produced by the 16 posterior blastomeres does not involve any asymmetric divisions.
Interestingly, in the Gls/Reg mutant, all of the cell
divisions are symmetric; the cells follow the ancestral Chlamydomonas pathway, acting first as vegetative cells and then re-differentiating into reproductive
cells. At the end of the seventh or eighth cell division
(which marks the end of the embryonic cleavage in
this mutant form that only has 128 or 256 cells), all
the cells are about 4 m in size. At the equivalent time
in development, the somatic initials in the wild-type
form are of the same size. However, due to the lesion
present in the regA gene, the Gls/Reg mutant cells are
able to grow post-embryonically and re-differentiate
into reproductive cells, whereas the wild-type somatic
cells, in which the regA gene is active, are destined to
remain small and terminally differentiated.
Thus, it is likely that under the scenario favored
by our model (Fig. 7B), subsequent to the seventh or
eighth symmetric cell division, the cells would fall
below the threshold size and would terminally differentiate into somatic cells; consequently, there would
be no reproductive cells to ensure the formation of
progeny. This problem could be resolved by having
the germ-line founder cells undergo fewer cell divisions, and by having more than one founder cells; for
instance, four instead of one germ-line founders that
would undergo a number of only two additional divisions would produce the same total number of 16
gonidia (4 × 22 ) and the critical cell size would not
be reached because the total number of cell divisions
would not exceed seven (5 + 2 = 7). However, the
embryo would be short of 384 somatic cells (3 × 27 =
111
384). Under the selective pressure to achieve a large
size and high motility, such a decrease in the total number of cells in the adult might be disfavored. Then, how
can the number of cells be increased while still producing 16 germ cells? Asymmetric divisions are able to do
that by keeping the size of one (i.e. the germ cell blastomeres) of the two daughter cells above the threshold,
while producing additional somatic blastomeres that
continue to proliferate below the threshold, and thus
increase the total number of somatic cells (Fig. 7A).
Therefore, the path that V. carteri undertook can be
explained as a consequence of the need to produce the
maximum number of germ cells and reach the largest
number of somatic cells, under the constraint of palintomy (i.e. no cell growth between cell divisions) and
the particular mechanism underlying somatic cell differentiation in this species (i.e. the expression of regA
when cell size falls below a threshold).
The other aspect that distinguishes the two alternatives for segregating a germ-line in V. carteri is concerned with the degree of relatedness among gonidia;
the gonidia are less related to each other under the
first scenario when compared to the second. It is possible that by having the 16 germ cells deriving from
16 independently segregated germ-line founder cells,
an increase in genetic variance among the progeny is
achieved. Because this species is reproducing mainly
asexually/clonally, such a pattern of producing the
germ cells might contribute to achieving genetic variation in populations. On the other hand, it has been
suggested that one of the advantages of having an early
segregating germ-line comes from the reduction in the
mutation level either by lowering the number of cell
divisions (Buss, 1987) or by lowering the mutation rate
per cell division (Michod, 1996, 1999; Michod and
Roze, 2000). It is interesting to note that the way that
the germ cells are produced in V. carteri ensures that
the deleterious mutations that might occur during the
last four cell divisions (i.e. the sixth, seventh, eighth,
and ninth) will be restricted to only one of the 16 germ
cells (Fig. 7A). In contrast, in the conventional way
of proliferating germ cells from only one germ-line
founder, deleterious mutations that would occur during the sixth division would be transmitted to half or
a quarter of the progeny (depending on whether the
mutation affects both or only one of the DNA strands,
respectively). In this light, the special way that the
germ-line is formed in V. carteri may be considered
112
R.E. Michod et al. / BioSystems 69 (2003) 95–114
as a means to allow the accumulation of non-lethal
variation, which is one of the premises for achieving
evolvability.
8. Germ-line, conflict mediation and evolvability
Although an early-segregated germ-line and a soma
have evolved in V. carteri, the soma was achieved in a
rather peculiar way. Volvox was not able to re-organize
immortality and totipotency among its cell lineages;
instead, in Volvox, both of these traits are entirely suppressed in the somatic cells. Moreover, the suppression of these traits was achieved by acting on a single
process, namely cell division. Furthermore, the way
in which Volvox suppressed cell division was not by
acting directly on the mitotic potential of the cells but
rather indirectly by acting on a trait that was still very
linked to it, that is the growth of the cell (Nedelcu
and Michod, 2003). By suppressing cell growth in somatic cells, cell division is repressed and the potential
for re-gaining immortality and totipotency (i.e. and for
gaining access to the germ-line) is “under control.” We
suggest that it is this type of conflict mediation affected
the potential for further evolution in this lineage.
A direct implication is that “soma” in Volvox differs
from the soma of other multicellular organisms. Because somatic cells do not divide, the post-embryonic
growth and/or regeneration of the individual are not
possible; in addition, because the somatic cells undergo senescence and cell death at the age of 5 days
(Pommerville and Kochert, 1981, 1982), the life span
of the higher-level individual is limited to the life
span of the lower-level somatic cell. Due to its unique
type of soma, Volvox is missing more than the ability to grow, regenerate, or live longer (whose lack
evidently does not constitute strong disadvantages in
the environment to which these algae are adapted,
namely temporal aquatic habitats). Without a mitotically active multipotent stem cell lineage or secondary
somatic differentiation there is less potential for cell
differentiation and further increases in complexity.
However, somatic growth and differentiation are important for the evolvability of a multicellular lineage.
Without them, Volvox did not and will likely not attain
higher-levels of complexity.
In conclusion, we have tried to understand how developmental processes are shaped during evolution-
ary transitions to increase the evolvability of the new
higher-level unit. In our formulation, evolvability of
the new unit of selection (multicellular organism) depends on enhanced cooperative interactions among
lower-level units (cells). Conflict mediation, the process by which cooperative interactions are enhanced
by reducing the temptation to defect, is instrumental in
creating a new evolutionary individual. Conflict mediation increases the cooperativity of the group and
heritability of fitness at the new level. Upon such processes does the continued evolvability of the new unit
evolutionary unit depend. Nevertheless, conflict mediation, like other selective processes, can be short-sited
and in the case of V. carteri appears to have interfered
with the long-term evolvability of the lineage.
Acknowledgements
We thank Laura Reed for discussions of modularity and evolvability and Lynne Trenery for reading
the manuscript and for discussions of the meaning of
cooperation.
Appendix A
The 2-locus 2-allele haploid dynamical system underlying this study has been introduced elsewhere
(Michod, 1997, 1999; Michod and Roze, 1997). In the
present study, we assume no recombination, r = 0.
We consider two loci, the first locus controls cell behavior (alleles C for cooperate and D for defect) and
the second modifier locus controls some aspect of
development or organization of the system (alleles m
for the ancestral state and M for the modified state).
Here the modifier allele M creates a sequestered
germ-line. There are four genotypes CM, Cm, DM,
and Dm, referred to as genotype 1–4, respectively, in
the subscript notation later.
In an organism with a germ-line (cell groups coming from a CM zygote) there is a distribution of C
and D cells at the time θ when the germ-line differentiates. We define k11,θ the number of CM cells at
the time θ, k31,θ the number of DM cells and k1,θ
the total number of cells (k1,θ = k11,θ , k31,θ ) coming
from a CM zygote. NC is the number of C cells in
the initial sample of ν cells which will give the germline and is given by a hypergeometric distribution
R.E. Michod et al. / BioSystems 69 (2003) 95–114
NC = H(k1,θ , ν, (k11,θ /k1,θ )). So we have for all i between 1 and ν,
k31,θ
k11,θ
ν−i
i
.
Pi = Pr(NC = i) =
k1,θ
ν
From our previous work (Michod, 1996; Michod
and Roze, 1997), we know the values of k11,θ , k31,θ ,
k1,θ at the time of the germ-line sample.
k11,θ = 2cθ (1 − µ)cθ ,
µ2bcθ − 2cθ (1 − µ)cθ µ
,
−1 + 2b−1 + µ
and k1,θ = k11,θ +k31,θ . Here µ is the mutation rate per
cell division from alleles C to D, c is the replication
rate of D cells (often assumed unity for simplicity and
without loss of generality) and cb is the replication rate
of mutant D cells. As we assume an infinite population,
we assume that Pi is the frequency of CM groups that
will give i C cells.
The ν cells in the germ-line sample go on and divide for a time tG to give the germ-line, and the mutation rate affecting the C allele is µG (a different rate,
perhaps, if the germ-line precursor cells are treated
differently). We call K11,i , K31,i and K1,i , the number
of C cells, the number of D cells and the total number
of cells in the germ-line of a CM individual whose
sample had i C cells. We have
k31,θ =
K11,i = i × 2ctG (1 − µG )ctG ,
K31,i = i
µG 2bctG − 2ctG (1 − µG )ctG µG
−1 + 2b−1 + µG
+(ν − i) × 2bctG ,
and k1,i = k11,i + k31,i .
We call k11,i , k31,i , and k1,i , the number of C cells,
the number of D cells and the total number of cells
in the soma of a CM individual whose sample for the
germ-line had i C cells. We then have
113
The expression of fitness for a group initiated by a
CM cell whose germ sample had i C cells is then
k11,i
,
W1,i = A 1 + β
k1,i
with A = 2ct (1−µ)ct +((µ2bct −2ct (1−µ)ct µ)/(−1+
2b−1 + µ)) − K1,i , if the cost of the germ-line is in
terms of the new germ cells (Fig. 3), and A = k1,i , if
the cost is in terms of number of missing somatic cells
(Fig. 4). Note that k1,i , is the number of somatic cells
after the germ-line has been sequestered and takes into
account the cells that are missing.
To date, we have only studied the case of asexual
reproduction, in which case the two-locus modifier
model has four equilibria (given and discussed in
Table 6–2 of Michod, 1999). The first equilibrium
and second equilibria do not interest us here as they
correspond to no cooperation (C allele in frequency
zero) with the modifier allele at frequency unity or
zero. The third and fourth equilibria correspond to
a population polymorphic at the cooperate/defect locus with the modifier M allele in frequency zero,
or unity, respectively. We are interested in the case
when the germ-line modifier allele M increases in
frequency and the system undergoes a transition from
equilibrium three to equilibrium four. This transition
corresponds to a transition in individuality, because
the group of cells are no longer indivisible, some
cells the somatic cells, can no longer produce the
group. Instability of equilibrium three is determined
by the following eigenvalue condition
(k22 /k2 )W2
< 1.
P
i=0 i (K11,i /K1,i )W1,i
λ = ν
If λ < 1, when the M allele appears in the population,
its frequency goes to fixation, and the system goes
from equilibrium three to equilibrium four. It is this
condition that leads to the results in Figs. 3, 4 and 8.
k11,i = 2ct (1 − µ)ct − i × 2c(t−θ) (1 − µ)c(t−θ) ,
k31,i =
µ2bct − 2ct (1 − µ)ct µ
−1 + 2b−1 + µ
µ2bc(t−θ) − 2c(t−θ) (1 − µ)c(t−θ) µ
−i
−1 + 2b−1 + µ
− (ν − i)2bc(t−θ) ,
and k1,i = k11,i + k31,i .
References
Anderson, R., Heasman, J., Wylie, C., 2001. Early events in the
mammalian germ-line. Int. J. Cytol. 203, 215–230.
Bell, G., 1985. The origin and early evolution of germ cells as
illustrated by the Volvocales. In: Halvorson, H.O., Monroy, A.
(Eds.), The Origin and Evolution of Sex, vol. 7. Alan R. Liss
Inc., New York, pp. 221–256.
114
R.E. Michod et al. / BioSystems 69 (2003) 95–114
Buss, L.W., 1987. The Evolution of Individuality. Princeton
University, Princeton, NJ.
Choi, G., Przybyiska, M., Straus, D., 1996. Three abundant germline-specific transcripts in Volvox carteri encode photosynthetic
proteins. Curr. Genet. 30, 347–355.
Desnitski, A.G., 1992. Cellular mechanisms of the evolution of
ontogenesis in Volvox. Archiv fur Protistenkunde 141, 171–178.
Desnitski, A.G., 1995. A review on the evolution of development
in Volvox—morphological and physiological aspects. Eur. J.
Protistol. 31, 241–247.
Drake, J.W., 1991. A constant rate of spontaneous mutation in
DNA-based microbes. Proc. Natl. Acad. Sci. U.S.A. 88, 7160–
7164.
Friedl, T., 1997. The evolution of green agae. Pl. Syst. Evol. 11,
87–101.
Huskey, R.J., Griffin, B.E., 1979. Genetic control of somatic cell
differentiation in Volvox. Dev. Biol. 72, 226–235.
Kirk, D.L., 1994. Germ cell specification in Volvox carteri.
In: Marsh, J., Goode, J. (Eds.), Proceedings of the Ciba
Symposium 184 on the Germ-line Development. Chichester,
Wiley, pp. 2–30.
Kirk, D.L., 1998. Volvox: Molecular-Genetic Origins of Multicellularity and Cellular Differentiation, 1st ed. Cambridge
University Press, Cambridge.
Kirk, D.L., 1999. Evolution of multicellularity in the volvocine
algae. Curr. Opin. Plant Biol. 2, 496–501.
Kirk, M.M., Ransick, A., Mcrae, S.E., Kirk, D.L., 1993. The
relationship between cell size and cell fate in Volvox carteri.
J. Cell Biol. 123, 191–208.
Kirk, M., Stark, K., Miller, S., Muller, W., Taillon, B., Gruber, H.,
Schmitt, R., Kirk, D.L., 1999. regA, a Volvox gene that plays
a central role in germ soma differentiation, encodes a novel
regulatory protein. Development 126, 639–647.
Kloc, M., Bilinski, S., Chan, A.P., Allen, L.H., Zearfoss,
N.R., Etkin, L.D., 2001. RNA localization and germ cell
determination in Xenopus. Int. J. Cytol. 203, 63–91.
Koufopanou, V., 1994. The evolution of soma in the Volvocales.
Am. Nat. 143, 907–931.
Koufopanou, V., Bell, G., 1991. Developmental mutants of Volvox:
does mutation recreate the patterns of phylogenetic diversity.
Evolution 45, 1806–1822.
Kurn, N., Colb, M., Shapiro, L., 1978. Spontaneous frequency of
a developmental mutant in Volvox. Dev. Biol. 66, 266–269.
Larson, A., Kirk, M.M., Kirk, D.L., 1992. Molecular phylogeny
of the volvocine flagellates. Mol. Biol. Evol. 9, 85–105.
Mahowald, A.P., 2001. Assembly of the Drosophila germ plasm.
Int. J. Cytol. 203, 187–213.
Margulis, L., 1981. Symbiosis in Cell Evolution. Freeman, San
Francisco.
Maynard Smith, J., Szathmáry, E., 1995. The Major Transitions
in Evolution. Freeman, San Francisco.
Meissner, M., Stark, K., Cresnar, B., Kirk, D.L., Schmitt, R.,
1999a. Volvox germ-line-specific genes that are putative targets
of RegA repression encode chloroplast proteins. Curr. Genet.
36, 363–370.
Meissner, M., Stark, K., Cresnar, B., Kirk, D.L., Schmitt, R.,
1999b. Volvox germ-line-specific genes that are putative targets
of RegA repression encode chloroplast proteins. Curr. Genet.
36, 363–370.
Michod, R.E., 1996. Cooperation and conflict in the evolution of
individuality. II. Conflict mediation. Proc. R. Soc. B 263, 813–
822.
Michod, R.E., 1997. Cooperation and conflict in the evolution of
individuality. I. Multi-level selection of the organism. Am. Nat.
149, 607–645.
Michod, R.E., 1999. Darwinian Dynamics, Evolutionary Transitions in Fitness and Individuality. Princeton University Press,
Princeton, NJ.
Michod, R.E., Nedelcu, A.M., 2003. Cooperation and conflict
during the unicellular–multicellular and prokaryotic–eukaryotic
transitions. In: Moya, A., Font, E. (Eds.), Evolution: From
Molecules to Ecosystems. Oxford University Press, Oxford.
Michod, R.E., Roze, D., 1997. Transitions in individuality. Proc.
R. Soc. B. 264, 853–857.
Michod, R.E., Roze, D., 1999. Cooperation and conflict in
the evolution of individuality. III. Transitions in the unit
of fitness. In: Nehaniv, C.L. (Ed.), Mathematical and Computational Biology: Computational Morphogenesis, Hierarchical Complexity, and Digital Evolution, vol. 26.
American Mathematical Society, Providence, Rhode Island,
pp. 47–92.
Michod, R.E., Roze, D., 2000. Some aspects of reproductive
mode and the origin of multicellularity. Selection 1, 97–
109.
Michod, R.E., Roze, D., 2001. Cooperation and conflict in the
evolution of multicellularity. Heredity 81, 1–7.
Miller, S.M., Kirk, D.L., 1999. GlsA, a Volvox gene required
for asymmetric division and germ cell specification, encodes a
chaperone-like protein. Development 126, 649–658.
Nakayama, T., Marin, B., Kranz, H.D., Surek, B., Huss, V.A.R.,
Inouye, I., Melkonian, M., 1998. The basal position of scaly
green flagellates among the green algae (chlorophyta) is
revealed by analyses of nuclear-encoded SSU rRNA sequences.
Protist 149, 367–380.
Nedelcu, A.M., Michod, R.E., 2003. Evolvability, modularity,
and individuality during the transition to multicellularity in
volvocalean green algae. In: Schlosser, G., Wagner, G. (Eds.),
Modularity in Development and Evolution. University Chicago
Press, Chicago.
Pommerville, J., Kochert, G., 1981. Changes in somatic cell
structure during senescence of Volvox carteri. Eur. J. Cell Biol.
24, 236–243.
Pommerville, J., Kochert, G., 1982. Effects of senescence on
somatic cell physiology in the green alga Volvox carteri.
Exp. Cell Res. 140, 39–45.
Roze, D., Michod, R.E., 2001. Mutation load, multi-level selection
and the evolution of propagule size during the origin of
multicellularity. Am. Nat. 158, 638–654.
Seydoux, G., Schedl, T., 2001. The germ-line in C. elegans: origins,
proliferation, and silencing. Int. J. Cytol. 203, 139–185.
Tam, L.W., Kirk, D.L., 1991. The program for cellular differentiation in Volvox carteri as revealed by molecular analysis
of development in a gonidialess/somatic regenerator mutant.
Development 112, 571–580.