Integrating "Evo" and "Devo": The Limb as Model Structure

Abstract

Reconciling the origins of morphological diversity with the deep homology of underlying mechanisms is a question fundamental to the goals of evolutionary developmental biology ("evo-devo" or EDB). In this paper I argue that differing research agendas in evolutionary and developmental biology have hindered how we address this question, but that the limb provides ideal "common ground" for their fuller integration. To support this idea, I review two previous analyses of limb variation in mammal, bird, and reptile taxa that offer complementary approaches to explaining diversity. Specifically, I present evidence suggesting that: (1) a shared genetic architecture affects the pattern of between limb developmental integration, while their functional dissociation is linked to both increased phenotypic evolvability and diversity of interlimb proportions, and (2) within limb proportional diversity is biased such that proximal and distal segments function as tradeoffs while the middle segment is more conservative, a signal that is both evident from early in morphogenesis and suggestive of an "inhibitory cascade" model of limb proximo-distal axis development. In the first case, shared genetic mechanisms predict both observed developmental integration between limbs and patterns of clade-specific diversity. In the second case, underappreciated patterns of phenotypic diversity suggest novel insights into the underlying developmental mechanisms by which variation is generated. These studies show how insights from both evolutionary and developmental biology of the limb may be used to generate novel testable hypotheses into the origins of diversity that are broadly applicable to the integration of EDB.

Full text

SYMPOSIUM
Integrating “Evo” and “Devo”: The Limb as Model Structure
Nathan M. Young
1
Department of Orthopaedic Surgery, University of California San Francisco, Building 9, Room 346, 2550 23rd Avenue,
San Francisco, CA 94110, USA
From the symposium “Physical and Genetic Mechanisms for Evolutionary Novelty” presented at the annual meeting of
the Society for Integrative and Comparative Biology, January 4–8, 2017 at New Orleans, Louisiana.
1
E-mail: nathan.young@ucsf.edu
Synopsis Reconciling the origins of morphological diversity with the deep homology of underlying mechanisms is a
question fundamental to the goals of evolutionary developmental biology (“evo–devo” or EDB). In this paper I argue
that differing research agendas in evolutionary and developmental biology have hindered how we address this question,
but that the limb provides ideal “common ground” for their fuller integration. To support this idea, I review two
previous analyses of limb variation in mammal, bird, and reptile taxa that offer complementary approaches to explaining
diversity. Specifically, I present evidence suggesting that: (1) a shared genetic architecture affects the pattern of between
limb developmental integration, while their functional dissociation is linked to both increased phenotypic evolvability
and diversity of interlimb proportions, and (2) within limb proportional diversity is biased such that proximal and distal
segments function as tradeoffs while the middle segment is more conservative, a signal that is both evident from early in
morphogenesis and suggestive of an “inhibitory cascade” model of limb proximo–distal axis development. In the first
case, shared genetic mechanisms predict both observed developmental integration between limbs and patterns of clade-
specific diversity. In the second case, underappreciated patterns of phenotypic diversity suggest novel insights into the
underlying developmental mechanisms by which variation is generated. These studies show how insights from both
evolutionary and developmental biology of the limb may be used to generate novel testable hypotheses into the origins of
diversity that are broadly applicable to the integration of EDB.
Integrating “Evo” and “Devo”
Genetic and molecular insights into the basis of mor-
phology over the previous three decades have driven a
re-appreciation for the role of development in evolu-
tion and the birth of a new field (evolutionary develop-
mental biology, “evo–devo”, or EDB) (e.g., see Raff
2000;Love 2003;Laublicher and Maienschein 2007;
Carroll et al. 2005;Moczek et al. 2015). With the re-
integration of these fields there have been hopes that,
among other goals (e.g., see Moczek et al. 2015), EDB
would grow into a science capable of predicting pat-
terns of phenotypic diversity from underlying genera-
tive mechanisms (Salazar-Ciudad and Jernvall 2004).
However, meaningful integration of EDB into a predic-
tive science arguably remains hindered by historical and
intellectual differences between the individual fields
(Love 2003). Moreover, there has been a tendency
toward “devo” over “evo” explanations, leading to calls
for a more balanced approach (Moczek 2012;
Diogo 2016). In this paper, I argue that to build a
more integrated EDB requires an appreciation of the
strengths of each discipline, a recognition of how these
strengths may inform model-building, and “common
ground” to build models.
One explanation for why EDB remains unbalanced
is that the kinds of questions evolutionary and devel-
opmental biologist traditionally address both require
different kinds of data and ultimately explain different
aspects of the same phenomena. To borrow Tinbergen’s
“levels of analysis” formulation (Tinbergen 1963;
MacDougall-Schackleton 2011), evolutionary biology
primarily focuses on “why” questions: e.g., why species
evolved over time (i.e., phylogeny), or why some traits
evolved over others (i.e., function or adaptation).
Evolutionary biology is also intrinsically focused on
variation, thus quantitative assessments of individual
differences at the population and species-level phe-
nomena are a major focus (e.g., Hallgr
ımsson and
ßThe Author 2017. Published by Oxford University Press on behalf of the Society for Integrative and Comparative Biology.
All rights reserved. For permissions please email: journals.permissions@oup.com.
Integrative and Comparative Biology
Integrative and Comparative Biology, volume 0, number 0, pp. 1–10
doi:10.1093/icb/icx115 Society for Integrative and Comparative Biology

Hall 2011). At the same time, the ability to quantify and
compare diversity has important explanatory limita-
tions. For example, multivariate methods are commonly
used to quantify and simplify complex phenotypes into
lower-dimensional “morphospaces” and often provide
important heuristic insight into biological phenomena
(Rohlf and Marcus 1993;Zelditch et al. 2012), but they
are often limited by what one observes (i.e., the data
input) and how one chooses to quantify it (Rohlf and
Marcus 1993;Zelditch et al. 1998).For even the simplest
organisms these choices are fraught with non-trivial
decisions (e.g., see Young 2005;Fitzhugh 2006;
Winther 2009). Consequently, the kinds of statistical
“top–down” approaches utilized in evolutionary biol-
ogy, while useful for characterizing diversity, are ulti-
mately mathematical constructs with no assurance that
they reflect how organismal morphology is constructed
(Rohlf and Marcus 1993;Zelditch et al. 1998).
In contrast, developmental biology has shifted in
concert with the molecular and genetic revolution to
a more “bottom up” and qualitative approach that pla-
ces greater value on proximate or mechanistic insight
(i.e., “how” questions) in the context of the individual
phenotype. Although individual phenotypes are under-
stood to be the realization of hierarchical, spatially and
temporally overlapping genetically encoded phenom-
ena (Hallgr

ımsson et al. 2009;Moczek et al. 2015), how
these phenomena generate quantitative phenotypic
variation remains largely unexplored. For example,
we often do not know how morphogenetic processes
may contribute to population-level differences (e.g.,
Atchley and Hall 1991), how developmental mecha-
nisms may vary among individuals or species, what
their potential to generate population-level variation
is, or how they might evolve. In practice, this means
that mechanistic data from developmental studies, of-
ten expressed as descriptive and qualitative outcomes,
may have limited utility when describing phenomena
that evolutionary biologists are interested in explaining
(although see Salazar-Ciudad and Jernvall 2004,2010;
Kavanagh et al. 2007).
This discussion suggests that both evolutionary and
developmental biology suffer from blind spots that a
more balanced and integrated field would ideally clar-
ify. While genetic insights gleaned from developmental
biology have already helped with “low-hanging fruit”
questions of macro-level scale evolution (e.g., Hox di-
versification and bodyplan evolution from flies to
humans) (Gellon and McGinnis 1998;Pearson et al.
2005), smaller scale questions that often dominate evo-
lutionary biology (e.g., between closely related species
or within populations) and their emphasis on quanti-
tative outcomes have proven more difficult to explain.
In this regard, the current focus of developmental
biology on mechanistic approaches to morphology
has the potential to be highly informative to measures
of diversity when coupled with explicit quantitative
modeling of both molecular inputs and phenotypic
outcomes (e.g., Salazar-Ciudad and Jernvall 2004,
2010;Kavanagh et al. 2007). Morphospaces constructed
from generative mechanisms of organismal morphol-
ogy (i.e., from the “bottom-up”) would be more bio-
logically grounded than purely statistical approaches,
and would have commensurately better ability to pre-
dict both realized and potential phenotypic diversity
(e.g., Salazar-Ciudad and Jernvall 2004,2010;
Matamoro-Vidal et al. 2015). At the same time, quan-
titative measures of diversity constructed from organ-
ismal phenotypes (e.g., morphospaces) can place
realistic “checks” on the construction of developmental
models by both demonstrating the limits of what is and
is not possible, and providing target data to directly test
predictions against (Fig. 1).
Six reasons to love the limb
Systems that are richly studied and well described
from the perspectives of both variation and mecha-
nism provide the ideal opportunity to derive and test
hypotheses about the generation of diversity at mul-
tiple scales using the principles outlined above. The
vertebrate limb arguably represents just such
Fig. 1 Complementary evidence from evolutionary and devel-
opmental biology. Variation is the principal focus of evolutionary
biology, yet how patterns of diversity are generated cannot be
fully addressed without knowledge of underlying generative
mechanisms. Developmental biology as a field focuses more on
deriving the proximate mechanisms driving morphogenesis, and
less on how generative capacities of these mechanisms influence
variation. Insights from patterns of observed diversity, for ex-
ample through the study of integration and modularity, may help
to inform models of development by predicting how variation
must be structured. In turn, modeling of developmental phe-
nomena may help to provide generative “rules” to help explain
how variation is generated and/or biased toward particular
outcomes.
2N. M. Young

“common ground” between disciplines, since it is a
particularly long-standing and compelling model
structure to address fundamental questions of how
development and evolution interact (e.g., Owen
1849;Huxley 1932;Oster et al. 1988). First, verte-
brate limb diversity presents remarkable contrasts
between the deep homology of anatomical organiza-
tion with functional and anatomical diversity (e.g.,
flying to swimming, quadrupeds to bipeds), consis-
tent with the question at the beginning of this paper
(Fig. 2). Second, there is a rich history of compara-
tive research of limbs that facilitates comparisons
and alleviates some concerns that groups are under-
represented (e.g., 2000þspecies-specific limb meas-
ures over the last 150 years) (Young 2013). Third,
limbs are serial homologs (Ruvinsky and Gibson-
Brown 2000;Adachi et al. 2016), meaning that there
is both functional diversification among species (e.g.,
quadrupeds versus bipeds) and limbs (e.g., forelimbs
versus hindlimbs) (Young and Hallgr
ımsson 2005),
and such structural independence limits the potential
confounding effects of direct physical interactions as
a source of covariation. Fourth, limbs share a com-
mon genetic architecture and developmental pro-
gram (Capdevila and Belmonte 2001). Fifth, limb
variation can be easily “simplified” to developmental
and functional components quantitatively defined by
their segment sizes, lengths, or proportions. Finally,
morphospaces that describe all potential combina-
tions of limbs can be constructed, even without un-
derstanding the generative mechanisms, and patterns
of diversity can be quantified and compared within
this morphospace, providing a useful feedback be-
tween data and model.
Below I review evidence from two studies that dem-
onstrate how this common ground can be transformed
into testable EDB models. In the first, information from
limb developmental genetics provides clues to observed
patterns of between limb variation and covariation
(Young and Hallgr
ımsson 2005;Young et al. 2010). In
the second, patterns of evolutionary limb diversity sug-
gest potential genetic mechanisms operating in limb
proximo–distal (PD) development and how they might
impact diversity (Young 2013;Young et al. 2015).
Between two limbs: what “devo” says
about “evo”
Limb evolution is a classic example of serial homology:
a primitive structure and its associated genetic archi-
tecture are duplicated in a new anatomical location,
subsequently diverging in both form and function
(Ruvinsky and Gibson-Brown 2000;Young and
Hallgr
ımsson 2005). The common origin of limbs as
serial homologs is evident from their shared genetic
architecture (e.g., Adachi et al. 2016), which further
implies a similar mechanistic logic driving their devel-
opmental similarities and differences. Because correla-
tions among traits are thought to be a product of
genetic pleiotropy (Wagner 1996), this shared architec-
ture predicts that limbs should exhibit stronger covari-
ation between homologous elements (e.g., humerus to
Fig. 2 The limb as a “model structure” in evo–devo. (A) Tetrapod limbs can be divided into a proximal stylopod [S], middle zeugopod
[Z], and distal autopod [A], itself divided into mesopod and acropod. This deep organizational homology reflects a conserved
underlying developmental program of proximo–distal (PD) axis formation. (B) Organizational homology is contrasted by proportional
diversity associated with functional differences. (C) Unlike other traits, limb segment proportions can be exhaustively described as
compositional data projected in a ternary morphospace (see Aitchison 1986;Comas-Cuf

ı et al. 2011), facilitating comparison of
observed diversity to predicted outcomes from developmental models. Cartoons show examples of limb segment proportions oc-
cupying different locations within the morphospace (shading as in 2A).
Integrating evo and devo 3

femur and tibia to ulna) than to non-homologous ele-
ments, all the more surprising given their structural
independence (Hallgr
ımsson et al. 2002).
My colleagues and I tested this hypothesis in two
studies, one looking at a broad set of functionally
diverse mammals (Young and Hallgr

ımsson 2005),
and a second focused on a narrower comparison
between radiations of suspensory hominoid apes (in-
cluding humans) and quadrupedal monkeys (Young
et al. 2010). Both studies found that there is stronger
covariation (i.e., integration) between homologous
limb segments, supporting the developmental model
(Fig. 3). Moreover, in both studies, the magnitude of
integration between limbs was dependent on the rel-
ative functional dissociation between forelimb and
hindlimb. For example, quadrupeds generally exhib-
ited the highest integration between limbs, consistent
with stronger selection on the coordination of opti-
mal between limb lengths, whereas species in which
the limbs are more functionally independent (e.g.,
suspensory apes, bipedal humans, and flying bats)
exhibited weaker correlations and overall integration.
This result implied that quadruped between-limb
proportions tend to be strongly coordinated, whereas
those of more functionally dissociated species are not
as well coordinated. This would further suggest that
as selection on divergent limb functions and associ-
ated proportions and lengths increases, the potential
for uncoordinated variation between limbs also
increases.
If relative covariation simply reflected the strength of
selection, then this result would be interesting but per-
haps less relevant for long-term evolution. However, if
covariation structure impacts the ability of a species to
respond to selection by limiting or uncovering varia-
tion (i.e., their relative “evolvability”) (Lande 1979;
Wagner et al. 2007), then relative integration may be
relevant to broader comparisons of patterns of diver-
sity. In other words, if strong covariation between traits
defines the evolutionary or adaptive “line of least
resistance” (sensu Schluter 1996), then variance in
some traits should be better coordinated in some direc-
tions but not others (Wagner et al. 2007). Strong inte-
gration would therefore imply a weak ability of
selection to evolve phenotypes orthogonal to this rela-
tionship, whereas weaker integration would enable se-
lection to independently evolve traits. Even in
situations where selection is not moving a species in a
particular direction in morphospace, the effect of drift
might be predicted to increase variation in more weakly
integrated groups since there is more space available to
“walk”.
These predictions are both directly derived from de-
velopmental insight and testable by comparing evolu-
tionary diversity in radiations of weakly integrated taxa
to more strongly integrated taxa (e.g., apes versus mon-
keys). In a range of measures of variability across dif-
ferent group structures, apes are significantly more
diverse in terms of between limb measures than are
quadrupedal monkeys (Fig. 4A). Moreover, if we
Fig. 3 Comparison of patterns of covariation and relative integration in the primate limb. (A) Primate limbs exhibit significant
correlations between developmentally homologous segments (H ¼humerus, R ¼radius, MC ¼metacarpal, F ¼femur, T ¼tibia,
MT ¼metatarsal). Strength of the shading is proportional to the strength of correlations (i.e., white corresponds to r¼0 and black to
r¼1). (B) Overall integration is lower in primates that have functionally dissociated limbs (i.e., suspensory apes) compared with
functionally associated quadrupeds, suggesting that selection directly impacts the coordination of variation across limbs. Data from
Young et al. (2010).
4N. M. Young

assume that functional dissociation is the primary cause
of lower integration then we might similarly predict that
“uncoordinated” limbs of non-quadrupeds are gener-
ally more diverse than those of “coordinated” quadru-
peds. Additional reanalysis of data published in Young
(2013)supports this idea, demonstrating that non-
quadrupeds are roughly two to three times more diverse
in their between limb proportions (Fig. 4B).
These results support the idea that shared devel-
opment impacts patterns of trait variation and
covariation in predictable ways, which in turn
impacts their evolvability, and ultimately broader
patterns of diversity. However, it is not meant to
imply that developmental integration imposes a strict
limit to variation, since selection on function likely
directly impacts how much variation is available.
Instead, these developmental relationships appear to
form a conserved baseline covariation of traits that
reflects their shared genetic architecture. For selec-
tion to operate in more strongly integrated traits,
Fig. 4 Comparison of functional dissociation to relative between limb diversity. (A) Radiations of highly suspensory apes (gibbons)
exhibit higher diversity of between limb proportions compared with quadrupedal primates (all others), even after controlling for similar
numbers of species and divergence time (Young et al. 2010). (B) Comparison of between species-level limb length data from Young
(2013)shows a similar pattern in which functional dissociation (and presumably lower integration) is associated with significantly
greater macroevolutionary diversity between limbs (Levene’s test, p<0.05), consistent with the limb developmental covariation model
(Hallgr

ımsson et al. 2002;Young and Hallgr

ımsson 2005;Young et al. 2010).
Integrating evo and devo 5

one would predict that first order changes would
require the breakdown of those pleiotropies under-
lying primitive patterns of covariation. Examples
from other serially homologous structures like but-
terfly wings in fact suggest that such changes are
both possible and amenable to strong directional se-
lection (Frankino et al. 2005,2007). What the ge-
netic targets of limb differentiation are in practice
remain largely unknown (i.e., beyond simply fore-
limb versus hindlimb [e.g., Takeuchi et al. 1999]),
but regulatory elements conferring specificity to
each limb during different developmental periods
of limb morphogenesis through postnatal growth
are likely targets (e.g., Cretekos et al. 2001;
Behringer et al. 2009;Rebeiz and Tsiantis 2017).
From patterns to mechanisms: what
“evo” says about “devo”
The contrast between the deep homology of tetrapod
limbs with their remarkable diversity in proportions
and functions is a common textbook biology figure,
and illustrates how limb diversity has a shared design
that spans both structural and genetic levels (see
Fig. 1A, B). While this is undoubtedly a powerful
example of evolution, it raises some equally important
questions about the nature of this diversity that are of-
ten overshadowed. For example: (1) are all proportions
equally possible, or are some combinations more likely
than others (i.e., is there bias to diversity?), and if so, (2)
is bias primarily driven by external factors such as func-
tion, internal factors such as development and morpho-
genesis, or some combination of both?
Limbs are particularly amenable to answering these
kinds of question for the reasons mentioned before:
they are both “simplify”-able and richly documented.
These facts were exploited to estimate how limb varia-
tion was apportioned among a broad sampling of am-
niote diversity (Young 2013). Contrary to unbiased a
priori predictions, variation in limb proportions is not
uniformly distributed across potential proportional
morphospace (Young 2013), instead almost all ob-
served combinations can be explained by PD segment
tradeoffs and a conservative middle segment (Fig. 5).
Moreover, this pattern does not seem to be an exclusive
function of growth, since it appears in early fetal pro-
portions of a diverse number of phylogenetically and
functionally divergent species (although, to be fair, this
possibility cannot be ruled out exhaustively for all taxa).
To be sure, there are notable exceptions, but the relative
Fig. 5 Observed amniote limb proportion diversity. Limb proportion data of individual species from Young (2013) is shown projected
into a ternary morphospace. Each side of the triangle represents the proportions of an individual element (stylopod ¼S, zeugopod ¼Z,
autopod ¼A) from 0 to 100% (see also Fig. 2C). Amniote limb proportions are not uniformly distributed across potential morpho-
space, and are notably under-represented by zeugopod diversity (note: autopod proportions are estimated from metapodial lengths,
which inflates both stylopod and zeugopod proportions). Proximal [S] and distal [A] proportional variation functions as a tradeoff that
accounts for the majority of covariation (gray shaded curve ¼data fit, black line ¼tradeoff model fit). Whether this tradeoff reflects
functional or developmental constraints is currently unknown.
6N. M. Young

rarity of divergent patterns underscores the predictive
value of the general rule. For example, plesiosaurs ex-
hibit an unusually small “carpalized” middle segment
(i.e., radius/ulna or tibia/fibula), due to a derived mech-
anism of perichondral ossification (Caldwell 1997).
Similar to irradiation experiments in which novel pro-
portions can be generated primarily by stunting indi-
vidual segmental growth (Summerbell 1981), the
plesiosaur “outlier” condition appears to derive from
altering later growth after early patterning events estab-
lish proportions. From these observations we can plau-
sibly hypothesize that a conserved PD patterning
mechanism helps to both establish segment propor-
tions early in amniote limb morphogenesis, and that
this mechanism is a target of selection (e.g., see
Sanger et al. 2012).
If species-specific proportions are both largely estab-
lished early in development and vary as a PD tradeoff
across functionally diverse species, then this would sug-
gest two non-exclusive possibilities. On the one hand,
middle segment proportional invariance may represent
a purely physical or biomechanical constraint. For ex-
ample, a relatively short or long middle segment pro-
portion might limit physical mobility, range of motion,
or biomechanics in ways that reduce functionality
across a wide range of adaptations and environments.
Alternatively, there could be a conserved mechanism of
limb morphogenesis that biases early middle segment
variability. If this signal appears early in development
(e.g., pre-fetal), it would indicate that patterning events
play an outsized role in determining species-specific
proportions. Indeed, if the signal was strong enough,
then one might also predict that it would be present
even despite the contribution of variation in later de-
velopmental events such as differential growth of seg-
ments. In either case, to my knowledge this pattern of
diversity has not been previously documented, so any
insight into how it is generated would represent a novel
insight into limb evolution. Below I focus primarily on
the second possibility, and explore what these compar-
ative insights suggest for how limb developmental
mechanisms may pattern the limb and how they are
modified to generate evolutionary variation.
How the PD axis is established in the limb is a classic
(and unresolved) question in developmental biology
(e.g., the “progress zone” [Wolpert 1969], “early spec-
ification” [Dudley et al. 2002], the “two-signal
gradient” [Mariani et al. 2008], and the “differentiation
front” [Tabin and Wolpert 2007]). It is also true that
most “models” are in fact post hoc explanations of ex-
perimental results, and are not explicitly designed to
explain or predict phenomena such as variation and/
or diversity. Moreover, implicit to the use of multiple
experimental organisms to dissect the mechanism is
that they are ultimately conserved, thus leaving unad-
dressed the question of how PD proportional diversity
is generated. Indeed, one could make the case that each
mechanism as described could be fine-tuned to indi-
vidual species, or in other words is both highly con-
served and evolvable (Young 2013). Alternatively, one
could argue that patterning is only important for estab-
lishing segment identities, whereas later developmental
events are where proportional variation of evolutionary
significance is generated, thus the contribution of pat-
terning to the larger question of diversity is moot.
My colleagues and I suggested that quantitative
comparative data might offer important clues to resolv-
ing the nature of the developmental mechanism behind
PD limb proportion generation and variation (Young
et al. 2015). We noted that, whereas PD models gener-
ally make no quantitative predictions about limb vari-
ation, observed diversity is strongly consistent with the
predictions of a generalized “inhibitory cascade” (IC)
model (Kavanagh et al. 2007). The IC model was orig-
inally proposed to model development and evolvability
of rodent molars and has been extended to a range of
species including humans (Evans et al. 2016). It would,
on first glance, seem an unlikely model for the limb (see
Newman and Mu¨ller [2005] for another non-canonical
take on PD limb development). However, experimental
evidence for the IC is also consistent with phalangeal
development (Kavanagh et al. 2013), suggesting the
possibility that similar dynamics are utilized in a range
of iteratively-forming structures or patterns (e.g.,
Kondo and Miura 2010).
In part, the generalizable nature of the IC model is a
product of its simple logic. In the model, the first seg-
ment forms in response to a proximal “activator” signal
that is in turn countered by a more distal flank inhib-
itory signal (Kavanagh et al. 2007). If we analogize this
process to limbs, Meis1 or RA would be candidates for
the activating flank signals that are balanced by lateral
inhibitory signals such as Fgfs(Mercader et al. 2000;
Mariani et al. 2008). Regardless of the actual signals,
the balance of the two generates segments both in iter-
ative fashion and with regularity in proportions, with
the middle segment always one-third of the total rela-
tive size. As a result, this model further predicts that PD
proportional variation is largely a function of proximal
and distal segment tradeoffs, consistent with the com-
parative data.
If early PD patterning events establish segments in
a “rule”-like fashion due to IC dynamics, then one
might predict that within population variation under
active selection would similarly show this kind of
signal. This appears to be true in forelimbs of the
domestic pigeon (Young et al. 2015), although anal-
ysis of similar data from extremes of human
Integrating evo and devo 7

proportions in Olympic athletes is more equivocal
(Young, unpublished data). Of course, a better test
would be to see whether one could select for longer
(or shorter) middle segment proportions than pre-
dicted by the model, and second to identify what
mechanisms drives these differences. For example,
stunted or altered post-patterning growth or dis-
rupted growth plates (as in irradiation experiments)
would not invalidate the model because they are al-
tering postnatal growth rather than the inferred lo-
cus of patterning. Evidence of species-specific
segment size post-patterning would be consistent
with the idea as well. However, generation of a range
of middle segment proportions early in development
would be strong evidence against this model.
Importantly, even if an IC-like PD mechanism in
limb patterning is falsified, the fact remains that middle
segments are more conservative than we might have
predicted a priori, implicating some unknown mecha-
nism. If middle segment proportions are shown to be
developmentally possible and thus evolvable beyond
what is typically observed in nature, then this begs the
question of what selective factor(s) drive this pattern
across such a broad sample of functional and phyloge-
netic diversity? In either case, it seems both evolution-
ary and developmental biologists have overlooked
critical patterns of diversity that must be accounted
for in any model attempting to explain the evolution
of vertebrate limbs.
Final thoughts
The primary goal of this paper was to highlight one
view about how the strengths of both evolutionary
and developmental biology can be utilized toward
building a more balanced field of EDB. Toward this
goal, I argued for the limb as a “model” structure for
probing fundamental EDB questions about the rela-
tionship between developmental conservation and phe-
notypic diversity. Such a proposal has deep precedent,
and certainly there are other structures that have equal
claim (e.g., teeth and vertebrae), but the value of the
limb is worth promoting. To support these ideas, I of-
fered two examples from my own work in which both
development and evolution suggest novel hypotheses
concerning the generation of diversity. Importantly,
even if the proposed mechanisms are ultimately falsi-
fied, the evidence of pattern presented here is robust,
and as of yet there is no satisfying alternative explana-
tions for how they are generated. I hope by highlighting
this approach to understanding patterns of limb diver-
sity, this discussion will inspire others to revisit the
questions raised here to build generalized and balanced
predictive EDB models.
Acknowledgments
I thank the organizers of the symposium, Thomas
Stewart, Stewart Newman, and Gu¨nter Wagner, for
inviting me to participate. Benedikt Hallgr
ımsson
contributed much of the intellectual mentorship for
my initial work on serial homology and integration
in limbs. Kathryn Kavanagh gamely collaborated on
expanding her novel “IC” model to less obvious con-
nections, like limbs and other segmented structures.
I also acknowledge my gratitude to Ralph Marcucio,
Theodore Miclau III, the members of the Laboratory
for Skeletal Regeneration, and the UCSF
Orthopaedic Trauma Institute for their ongoing ma-
terial support and intellectual encouragement.
Funding
Many of the ideas expressed herein had their initial
genesis during a postdoctoral grant funded by
Alberta Ingenuity [No. 200300516]. Rui Diogo and
two anonymous reviewers provided useful feedback.
References
Adachi N, Robinson M, Goolsbee A, Shubin NH. 2016.
Regulatory evolution of Tbx5 and the origin of paired
appendages. Proc Natl Acad Sci U S A 113:10115–20.
Aitchison J. 1986. The statistical analysis of compositional
data. London: Chapman and Hall.
Atchley WR, Hall BK. 1991. A model for development and
evolution of complex morphological structures. Biol Rev
66:101–57.
Behringer RR, Rasweiler JJ 4th, Chen CH, Cretekos CJ. 2009.
Genetic regulation of mammalian diversity. Cold Spring
Harb Symp Quant Biol 74:297–302.
Caldwell MW. 1997. Modified perichondral ossification and
the evolution of paddle-like limbs in ichthyosaurs and ple-
siosaurs. J Vert Paleontol 17:534–47.
Capdevila J, Belmonte JCI. 2001. Patterning mechanisms con-
trolling vertebrate limb development. Ann Rev Cell Dev
Biol 17:87–132.
Carroll SB, Grenier JK, Weatherbee SD. 2005. From DNA to
diversity: molecular genetics and the evolution of animal
design. 2nd ed. Malden (MA): Blackwell Publishing Ltd.
Comas-Cuf
ıM,Thi

o-Henestrosa S, Egozcue JJ, Tolosana-
DelgadoR,OrtegoMI.2011.CoDaWork’11:4th
International Workshop on Compositional Data Analysis.
Sant Feliu de Gu

ıxols.
Cretekos CJ, Rasweiler JJ, Behringer RR. 2001. Comparative
studies on limb morphogenesis in mice and bats: a func-
tional genetic approach toward a molecular understanding
of diversity in organ formation. Reprod Fertil Dev
13:691–5.
Diogo R. 2016. Where is the evo in evo–devo (evolutionary
developmental biology)? J Exp Zool B Mol Dev Evol
326B:9–18.
8N. M. Young

Dudley AT, Ros MA, Tabin CJ. 2002. A re-examination of
proximodistal patterning during vertebrate limb develop-
ment. Nature 418:539–44.
Evans AR, Daly ES, Catlett KK, Paul KS, King SJ, Skinner
MM, Nesse HP, Hublin JJ, Townsend GC, Schwartz GT,
et al. 2016. A simple rule governs the evolution and devel-
opment of hominin tooth size. Nature 530:477–80.
Fitzhugh K. 2006. The philosophical basis of character coding for
the inference of phylogenetic hypotheses. Zool Scr 35:261–86.
Frankino WA, Zwaan BJ, Stern DL, Brakefield PM. 2005.
Natural selection and developmental constraints in the evo-
lution of allometries. Science 307:718–20.
Frankino WA, Zwaan BJ, Stern DL, Brakefield PM. 2007.
Internal and external constraints in the evolution of mor-
phological allometries in a butterfly. Evolution 61:2958–70.
Gellon G, McGinnis W. 1998. Shaping animal body plans in
development and evolution by modulation of Hox expres-
sion patterns. BioEssays 20:116–25.
Hallgr
ımsson B, Hall BK. 2011. Variation: a central concept
in biology. New York (NY): Academic Press.
Hallgr
ımsson B, Willmore K, Hall BK. 2002. Canalization,
developmental stability, and morphological integration in
primate limbs. Am J Phys Anthropol 45:131–58.
Hallgr
ımsson B, Jamniczky H, Young NM, Rolian C, Parsons
TE, Boughner JC, Marcucio RS. 2009. Deciphering
the palimpsest: studying the relationship between morphologi-
cal integration and phenotypic covariation. Evol Biol 36:355–76.
Huxley JS. 1932. Problems of relative growth. London:
Methuen & Company.
Kavanagh K, Evans A, Jernvall J. 2007. Predicting evolution-
ary patterns of mammalian teeth from development.
Nature 449:427–32.
Kavanagh K, Shoval O, Winslow B, Alon U, Leary B, Kan A,
Tabin CJ. 2013. Developmental bias in the evolution of
phalanges. Proc Natl Acad Sci U S A 110:18190–5.
Kondo S, Miura T. 2010. Reaction-diffusion model as a
framework for understanding biological pattern formation.
Science 329:1616–20.
Lande R. 1979. Quantitative genetic analysis of multivariate evolu-
tion, applied to brain:body size allometry. Evolution 33:402–16.
Laublicher MD, Maienschein J. 2007. From embryology to
evo–devo. Cambridge: MIT Press.
Love AC. 2003. Evolutionary morphology, innovation, and
the synthesis of evolutionary and developmental biology.
Biol Philos 18:309–45.
MacDougall-Schackleton SA. 2011. The levels of analysis
revisited. Phil Trans R Soc B 366:2076–85.
Mariani FV, Ahn CP, Martin GR. 2008. Genetic evidence that
FGFs have an instructive role in limb proximal–distal pat-
terning. Nature 453:401–5.
Matamoro-Vidal A, Salazar Ciudad I, Houle D. 2015. Making
quantitative variation from basic developmental processes:
where are we? The case of the Drosophila wing. Dev Dyn
244:1058–73.
Mercader N, Leonardo E, Piedra ME, Mart
ınez-A C, Ros MA,
Torres M. 2000. Opposing RA and FGF signals control
proximodistal vertebrate limb development through regu-
lation of Meis genes. Development 127:3961–70.
Moczek AP. 2012. The nature of nurture and the future of
evodevo: toward a theory of developmental evolution.
Integr Comp Biol 52:108–19.
Moczek AP, Sears KE, Stollewerk A, Wittkopp PJ, Diggle P,
Dworkin I, Ledon-Rettig C, Matus DQ, Roth S, Abouheif
E, et al. 2015. The significance and scope of evolutionary
developmental biology: a vision for the 21st century. Evol
Dev 17:198–219.
Newman SA, Mu¨ ller GB. 2005. Origination and innovation in
the vertebrate limb skeleton: an epigenetic perspective.
J Exp Zool B Mol Dev Evol 304B:593–609.
Oster GF, Shubin N, Murray JD, Alberch P. 1988. Evolution
and morphogenetic rules: the shape of the vertebrate limb
in ontogeny and phylogeny. Evolution 42:862–84.
Owen R. 1849. On the nature of limbs. London: John Van Voorst.
Pearson JC, Lemons D, McGinnis W. 2005. Modulating Hox
gene functions during animal body patterning. Nat Rev
Genet 6:893–904.
Raff RA. 2000. Evo–devo: the evolution of a new discipline.
Nat Rev Genet 1:74–9.
Rebeiz M, Tsiantis M. 2017. Enhancer evolution and the origins
of morphological novelty. Curr Opin Genet Dev 45:115–23.
Rohlf FJ, Marcus LF. 1993. A revolution in morphometrics.
Trends Ecol Evol 8:129–32.
Ruvinsky I, Gibson-Brown JJ. 2000. Genetic and developmen-
tal bases of serial homology in vertebrate limb evolution.
Development 127:5233–44.
Salazar-Ciudad I, Jernvall J. 2004. How different types of pat-
tern formation mechanisms affect the evolution of form
and development. Evol Dev 6:6–16.
Salazar-Ciudad I, Jernvall J. 2010. A computational model of
teeth and the developmental origins of morphological var-
iation. Nature 464:583–6.
Sanger TJ, Revell LJ, Gibson-Brown JJ, Losos JB. 2012.
Repeated modification of early limb morphogenesis pro-
grammes underlies the convergence of relative limb length
in Anolis lizards. Proc Biol Sci 279:739–48.
Schluter D. 1996. Adaptive radiation along genetic lines of
least resistance. Evolution 50:1766–74.
Summerbell D. 1981. Evidence for regulation of growth, size
and pattern in the developing chick limb bud. J Embryol
Exp Morphol 65:129–50.
Tabin C, Wolpert L. 2007. Rethinking the proximodistal axis of the
vertebrate limb in the molecular era. Genes Dev 21:1433–42.
Takeuchi JK, Koshiba-Takeuchi K, Matsumoto K, Vogel-
Hopker A. 1999. Tbx5 and Tbx4 genes determine the
wing/leg identity of limb buds. Nature 398:810.
Tinbergen N. 1963. On aims and methods of ethology.
Ethology 20:410–33.
Wagner GP. 1996. Homologues, natural kinds and the evo-
lution of modularity. Am Zool 36:36–43.
Wagner GP, Pavlicev M, Cheverud JM. 2007. The road to
modularity. Nat Rev Genet 8:921–31.
Winther RG. 2009. Character analysis in cladistics: abstrac-
tion, reification, and the search for objectivity. Acta
Biotheor 57:129–62.
Wolpert L. 1969. Positional information and the spatial pat-
tern of cellular differentiation. J Theor Biol 25:1–47.
Young NM. 2005. Estimating hominoid phylogeny from mor-
phological data: character choice, phylogenetic signal and
postcranial data. In: Lieberman DE, Smith RJ, Kelley J,
editors. Interpreting the past: essays on human, primate
and mammal evolution in honor of David Pilbeam.
Boston: Brill Academic Press. p. 19–31.
Integrating evo and devo 9

Young NM, Hallgr
ımsson B. 2005. Serial homology and the
evolution of mammalian limb covariation structure.
Evolution 59:2691–704.
Young NM, Wagner GP, Hallgr
ımsson B. 2010. Development
and the evolvability of human limbs. Proc Natl Acad Sci U
S A 107:3400–5.
Young NM. 2013. Macroevolutionary diversity of amniote
limb proportions predicted by developmental interactions.
J Exp Zool B Mol Dev Evol 320:420–7.
Young NM, Winslow B, Takkellapati S, Kavanagh K. 2015.
Shared rules of development predict patterns of evolution
in vertebrate segmentation. Nat Commun 6:6690.
Zelditch ML, Fink WL, Swiderski DL, Lundrigan DL. 1998.
On applications of geometric morphometrics to studies of
ontogeny and phylogeny: a reply to Rohlf. Syst Biol 47:159–67.
Zelditch ML, Swiderski DL, Sheets HD. 2012. Geometric
morphometrics for biologists: a primer. New York (NY):
Academic Press.
10 N. M. Young