A genetic signature of the evolution of loss of flight in the Galapagos cormorant

Abstract

Loss of flight in the Galapagos cormorant
Although rare among existing birds, the loss of flight appears to have occurred multiple times in evolutionary history. However, the genetic changes that ground avian species are not well understood. Burga et al . sequenced genomes from three cormorant species and compared them with that of the flightless Galapagos cormorant (see the Perspective by Cooper). They identified variants in genes involved in primary ciliogenesis. Functional analyses of these variants suggest that the impaired function of the genes may be responsible for skeletal changes associated with the Galapagos cormorant’s loss of flight.
Science , this issue p. eaal3345 ; see also p. 904

Full text

RESEARCH ARTICLE SUMMARY
◥
EVOLUTIONARY GENOMICS
A genetic signature of the
evolution of loss of flight in
the Galapagos cormorant
Alejandro Burga,*Weiguang Wang, Eyal Ben-David, Paul C. Wolf, Andrew M. Ramey,
Claudio Verdugo, Karen Lyons, Patricia G. Parker, Leonid Kruglyak*
INTRODUCTION: Changes in the size and pro-
portion of limbs and other structures have
played a key role in the evolution of species.
One common class of limb modification is re-
current wing reduction and loss of flight in
birds. Indeed, Darwin used the occurrence of
flightless birds as an argument in favor of his
theory of natural selection. Loss of flight has
evolved repeatedly and is found among 26 fam-
ilies of birds in 17 different orders. Despite the
frequency of these modifications, we have a
limited understanding of their underpinnings
at the genetic and molecular levels.
RATIONALE: To better understand the evolu-
tion of changes in limb size, we studied a classic
case of recent loss of flight in the Galapagos
cormorant (Phalacrocorax harrisi). Cormorants
are large water birds that live in coastal areas
or near lakes, and P. harrisi istheonlyflight-
less cormorant among approximately 40 extant
species. The entire population is distributed
along the coastlines of Isabela and Fernandina
islands in the Galapagos archipelago. P. harrisi
has a pair of short wings, which are smaller
than those of any other cormorant. The extreme
reduction of the wings and pectoral skeleton
observed in P. harrisi is an attractive model
for studying the evolution of loss of flight be-
cause it occurred very recently; phylogenetic evi-
dence suggests that P. harrisi diverged from its
flighted relatives within the past 2 million years.
We developed a comparative and predictive ge-
nomics approach that uses the genome se-
quences of P. harrisi and its flighted relatives
to find candidate genetic variants that likely
contributed to the evolution of loss of flight.
RESULTS: We sequenced and de novo assem-
bled the whole genomes of P. harrisi and three
closely related flighted cormorant species. We
identified thousands of coding variants exclu-
sive to P. harrisi andclassifiedthemaccording
to their probability of altering protein function
based on conservation. Variants most likely to
alter protein function were significantly enriched
in genes mutated in human skeletal ciliopathies,
including Ofd1,Evc,Wdr34,andIft122.Wecar-
ried out experiments in Caenorhabditis elegans
to confirm that a missense variant present in
the Galapagos cormorant IFT122 protein is
sufficient to affect ciliary function. The pri-
mary cilium is essential for Hedgehog (Hh)
signaling in vertebrates, and individuals af-
fected by ciliopathies have small limbs and
ribcages, mirroring the phenotype of P. harrisi.
We also identified a 4–amino acid deletion in
theregulatorydomainofCux1,ahighlycon-
served transcription factor
that has been expe rim ent -
ally shown to regulate limb
growth in chicken. The four
missing amino acids are
perfectly conserved in all
birds and mammals se-
quenced to date. We tested the consequences
of this deletion in a chondrogenic cell line and
showed that it impairs the ability of CUX1 to
transcriptionally up-regulate cilia-related genes
(some of which contain function-altering var-
iants in P. harrisi) and to promote chondrogenic
differentiation. Finally, we show that positive
selection may have played a role in the fixation
of the variants associated with loss of flight in
P. harrisi.
CONCLUSION: Our results indicate that the
combined effect of variants in genes necessary
for the correct transcriptional regulation and
function of the primary cilium likely contrib-
uted to the evolution of highly reduced wings
and other skeletal adaptations associated with
loss of flight in P. harrisi. Our approach may
be generally useful for identification of var-
iants underlying evolutionary novelty from ge-
nomes of closely related species. ▪
RESEARCH
Burga et al., Science 356 , 921 (2017) 2 June 2017 1of 1
The list of author affiliations is available in the full article online.
*Corresponding author. Email: aburga@mednet.ucla.edu
(A.B.); lkruglyak@mednet.ucla.edu (L.K.)
Cite this article as A. Burga et al., Science 356, eaal3345
(2017). DOI: 10.1126/science.aal3345
Comparative and predictive genomics of loss of flight. Comparison of the genomes of four closely related cormorant species allowed us
to predict function-altering variants exclusively affecting the Galapagos cormorant and to test their functional consequences. Our results
implicate ciliary dysfunction as a likely contributor to the evolution of loss of flight.
ON OUR WEBSITE
◥
Read the full article
at http://dx.doi.
org/10.1126/
science.aal3345
..................................................
on June 13, 2017http://science.sciencemag.org/Downloaded from

RESEARCH ARTICLE
◥
EVOLUTIONARY GENOMICS
A genetic signature of the
evolution of loss of flight in
the Galapagos cormorant
Alejandro Burga,
1
*Weiguang Wang,
2
Eyal Ben-David,
1
Paul C. Wolf,
3
Andrew M. Ramey,
4
Claudio Verdugo,
5
Karen Lyons,
2
Patricia G. Parker,
6,7
Leonid Kruglyak
1
*
We have a limited understanding of the genetic and molecular basis of evolutionary
changes in the size and proportion of limbs. We studied wing and pectoral skeleton
reduction leading to flightlessness in the Galapagos cormorant (Phalacrocorax harrisi).
We sequenced and de novo assembled the genomes of four cormorant species and applied
a predictive and comparative genomics approach to find candidate variants that may
have contributed to the evolution of flightlessness. These analyses and cross-species
experiments in Caenorhabditis elegans and in chondrogenic cell lines implicated variants
in genes necessary for transcriptional regulation and function of the primary cilium.
Cilia are essential for Hedgehog signaling, and humans affected by skeletal ciliopathies
suffer from premature bone growth arrest, mirroring skeletal features associated with
loss of flight.
The evolution of loss of flight is one the most
recurrent limb modifications encountered
in nature (1). Indeed, Darwin used the occur-
rence of flightless birds as an argument in
favor of his theory of natural selection (2).
He proposed that loss of flight could evolve as a
result of selection in favor of larger bodies and
relaxed selection due to the absence of predators.
Loss of flight has evolved repeatedly and is found
among 26 families of birds in 17 different orders
(1). Moreover, recentstudies strongly suggest that
the ratites (ostriches, emus, rheas, cassowaries,
and kiwis), long thought to derive from a single
flightless ancestor, may constitute a polyphyletic
group characterized by multiple independent in-
stances of loss of flight and convergent evolution
(3–5). However, despite the ubiquity and evolu-
tionary importance of loss of flight (6), the under-
lying genetic and molecular mechanisms remain
unknown.
The Galapagos cormorant (Phalacrocorax harrisi)
istheonlyflightlesscormorantamongapproxi-
mately 40 extant species (7).Theentirepopulation
is distributed along the coastlines of Isabela
and Fernandina islands in the Galapagos archi-
pelago. P. harrisi has a pair of short wings, which
are smaller than those of any other cormorant
(Fig. 1A)—a deviation from the allometric relation-
ship between wing length and body mass (7). The
radius and ulna are disproportionately small rela-
tive to the humerus, but no digits have been fused
or lost, unlike in some ratites (8). In addition, the
Galapagos cormorant differs from its flighted rela-
tives in a delay in the onset of several develop-
mental landmarks after hatching (9), shortened
remiges (flight feathers), underdeveloped pectoral
muscles, a long and narrow skull and pelvis, a dis-
proportionately long tibiotarsus, a factor of 1.6 in-
crease in body mass, and a highly reduced keel (7).
The keel is an extension of the sternum that runs
along its midline and provides an attachment sur-
face for the flight muscles, the largest muscles in
birds. Flightless taxa, such as ratites and Cretaceous
Hesperornis, have evolved flat sternums in which
the keel has been largely reduced or lost (10).
In contrast to ratites and penguins, which be-
came flightless more than 50 million years ago
(Ma) (5,11), the Galapagos cormorant and its flighted
relatives are estimated to share a common ancestor
at ~2 Ma (12). This recent and extreme modification
of wing size and pectoral skeleton makes P. harrisi
an attractive model for studying loss of flight.
High-quality genome sequences of four
cormorant species
To identify variants associated with loss of flight,
we sequenced and de novo assembled the 1.2-Gb
genomes of the Galapagos cormorant (Galapagos
Islands, Ecuador) and three flighted cormorant spe-
cies: the double-crested cormorant (Phalacrocorax
auritus; Minnesota, USA), the neotropical cormo-
rant (Phalacrocorax brasilianus; Valdivia, Chile),
and the pelagic cormorant (Phalacrocorax pelagicus;
Alaska, USA). P. auritus and P. brasilianus are the
closest relatives of P. harrisi (12–14), and P. pelagicus
ispartofasistercladeandservedasanoutgroup.
Genomes were assembled from a combination of
short insert and mate-pair Illumina libraries with
SOAPdenovo2 (15) (table S1). Among these four
genomes, the Galapagos cormorant’sassembly
had the longest contig and scaffold N50 metrics
(contig N50, 103 kb; scaffold N50, 4.6 Mb; table
S1B). We evaluated the completeness of the cor-
morants’genomesbyestimatingthetotalnum-
ber of uniquely annotated proteins in each assembly
andbyusingtheCEGMApipeline(16,17). Overall,
we found agreement between these two indepen-
dent metrics in a data set including the four cor-
morant genomes and 17 recently published bird
genomes (r
2
=0.75,P=4.3×10
−7
;Fig.1Band
table S2).
Commonly used metrics of assembly quality,
such as contig and scaffold N50, were very poor
predictors of the total number of proteins present
in each assembly (r
2
=0.13,P=0.15,andr
2
=
0.06, P= 0.81; fig. S1 and table S2). Three of the
four cormorant genomes (P. harrisi,P. auritus,
and P. pelagicus) obtained the highest CEGMA
scores and numbers of uniquely annotated genes
among all bird genomes (red triangles, Fig. 1B).
The following CEGMA scores, a means to estimate
genome completeness, were obtained for the cor-
morants: P. harrisi,90.3%;P. auritus,91.3%;
P. brasilianus,72.6%;P. pelagicus,87.1%.Incon-
trast, Sanger and PacBio genomes had lower scores
for other birds: Gallus gallus (Sanger assembly)
(18), 80.7%; Taeniopygia guttata (Sanger assembly)
(19), 71.4%; and Melopsittacus undulatus (PacBio
assembly) (20), 79.0%. Thus, our cormorant ge-
nomes perform even better than genomes assem-
bled from Sanger sequences and PacBio long
reads (complete statistics in table S2).
Phylogeny and genetic diversity
We reconstructed the cormorant phylogeny using
a Bayesian framework (17) and confirmed the phy-
logenetic relationship among the four sequenced
species (Fig. 1C). Moreover, our results indicate
that P. harrisi last shared a common ancestor with
P. auritus and P. brasilianus at ~2.37 Ma, in
agreement with an estimate from mitochondrial
DNA (12) (Fig. 1C). Española, the oldest extant
island in the Galapagos archipelago, emerged no
earlier than 4 Ma, and proto-Galapagos islands
existed at 9 Ma or earlier (21). Our results are con-
sistent with the view that P. harrisi lost the abil-
ity to fly while inhabiting the archipelago.
We calculated the proportion of single-nucleotide
polymorphism (SNP) heterozygous sites for each
sequenced individual to estimate the levels of in-
traspecific genetic diversity (Fig. 1D). P. harrisi
showed the lowest proportion of heterozygous
SNPs among the sequenced cormorants (0.00685%;
Fig. 1D). The heterozygosity of P. harrisi is even
lower than that of the crested ibis, Nipponia nippon,
a highly endangered bird with a small effective
RESEARCH
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 1of8
1
Department of Human Genetics, Department of Biological
Chemistry, and Howard Hughes Medical Institute, University
of California, Los Angeles, CA, USA.
2
Departments of
Molecular, Cell and Developmental Biology and Orthopaedic
Surgery, University of California and Orthopaedic Institute for
Children, Los Angeles, CA, USA.
3
Wildlife Services, U.S.
Department of Agriculture, Roseburg, OR, USA.
4
U.S. Geological
Survey Alaska Science Center, Anchorage, AK, USA.
5
Instituto de Patología Animal, Facultad de Ciencias
Veterinarias, Universidad Austral de Chile, Valdivia, Chile.
6
Department of Biology and Whitney Harris World Ecology
Center, University of Missouri, St. Louis, MO, USA.
7
WildCare
Institute, Saint Louis Zoo, St. Louis, MO, USA.
*Corresponding author. Email: aburga@mednet.ucla.edu (A.B.);
lkruglyak@mednet.ucla.edu (L.K.)
on June 13, 2017http://science.sciencemag.org/Downloaded from

population size and a known recent population
bottleneck (22) (0.0172%; Fig. 1D). The low level
of heterozygosity found intheGalapagoscormo-
rant is most likely due to its small population
size (~1500 individuals) and multiple popula-
tion bottlenecks (23).
Discovery and characterization of
function-altering variants in P. harrisi
To investigate the genetics of flightlessness evo-
lution, we developed a comparative and predic-
tive genomics approach (24,25) that uses the
genome sequences of P. harrisi and its flighted
relatives to identify genetic variants that likely
contributed to the evolution of loss of flight. Both
coding and cis-regulatory variants have been im-
plicated in the evolution of morphological traits
(26,27). However, determining the impact of
regulatory variants is not straightforward. To iden-
tify the contribution of regulatory variants to the
evolutionoflossofflightinP. harrisi,wesearched
for ultraconserved noncoding sequences showing
accelerated molecular evolution (28–30). We iden-
tified 11 ultraconserved noncoding regions in tetra-
pods that show accelerated evolution in P. harrisi
but not in the other cormorants [false discovery
rate (FDR) < 5%]. One of these regions was lo-
cated in an intron of the gene FTO (fig. S2), which
has been associated with obesity in humans (31);
however, none of these regions overlapped with
experimentally validated or putative mouse limb
enhancers (17,32,33)(tableS7).
We thus focused on characterizing coding var-
iants because we are better able to predict their
molecular consequences. For our variant discovery
approach to be comprehensive, it was imperative
to interrogate most of the Galapagos cormorant’s
genes. To increase our power to do so, we annotated
genes using homology-based and transcriptome-
based gene annotations. The latter were derived
with mRNA expression data from the developing
wing of a double-crested cormorant embryo (fig.
S3C) (17). We then predicted all missense, dele-
tion, and insertion variants in ortholog pairs be-
tween P. harrisi and each of its three flighted
relatives (fig. S4) (17).
We used PROVEAN (34), a phylogeny-corrected
variant effect predictor, to evaluate the impact on
proteinfunctionofeachoftheGalapagoscormo-
rant’svariantsonagenome-widescale.PROVEAN
predictions have been validated in experimental
evolution studies that mimic the process of grad-
ual accumulation of mutations in nature (35). A
PROVEAN score is calculated for each variant;
the more negative the score, the more likely a
given variant is to alter protein function. We ex-
amined the distribution of PROVEAN scores ob-
tained when comparing 12,442 ortholog pairs
between P. harrisi and P. auritus (Fig. 2A). Of
these, 4959 pairs (40%) did not contain coding
variants; the remaining 7483 pairs contained a
total of 23,402 coding variants: 22,643 single
amino acid substitutions, 456 deletions, and 303
insertions (Fig. 2B). Most variants were predicted
to be neutral (the distribution is centered around
zero). As expected, deletions and insertions were
enriched in the tails of the distribution (Fig. 2B).
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 2of8
Fig. 1. The Galapagos cormorant, a model for studying flightlessness evolution. (A) The
average wing length of an adult Galapagos cormorant male is 19 cm (3.6 kg body mass), whereas the
wing length of its closest relative, the double-crested cormorant, is 31.5 cm (2.2 kg body mass).
[Illustration by Katie Bertsche from specimens 134079 and 151575, Museum of Vertebrate Zoology at
Berkeley] (B) The CEGMA score is a good predictor of genome completeness from a gene-centric
perspective. The blue line is the linear regression model (r
2
=0.75,P=4.3×10
−7
). Genomes
reported in this study are red triangles; other published avian genomes are black circles (table S2).
(C) Bayesian phylogram reconstructed with fourfold degenerate sites from whole genome
sequences. The orange bar illustrates the time span between the approximate origin of the proto-
Galapagos archipelago (9 Ma) and the origin of the oldest extant island, San Cristobal (4 Ma). Nodes
represent median divergence ages. Blue bars indicate the 95% highest posterior density interval.
(D) Heterozygosity levels inferred from whole genome sequences. Birds are not drawn to scale.
Table 1. Function-altering variants in P. harrisi are enriched for genes that cause skeletal
ciliopathies in humans. Sanger-validated examples of function-altering variants (PROVEAN score
<–5) in P. harrisi. Cilia/Hh-related genes were found on the basis of functional enrichment for
human syndromes. PCP (planar cell polarity) genes were selected according to literature evidence
linking cilia and PCP. These variants are fixed in the population.
Gene Pathway Variant PROVEAN score Human syndrome
Ofd1 Cilia/Hh R325C
K517T
E889G
–6.913
–5.673
–5.068
Orofaciodigital and Joubert
................................... ................................................................ ............................................................... ....................................................
Talpid3 Cilia/Hh D759V –7.805 Joubert and Jeune
................................... ................................................................ ............................................................... ....................................................
Evc Cilia/Hh T341I –5.546 Ellis–van Creveld
................................... ................................................................ ............................................................... ....................................................
Dync2h1 Cilia/Hh P2733S –7.431 Short-rib thoracic dysplasia
................................... ................................................................ ............................................................... ....................................................
Ift122 Cilia/Hh Q691L –5.491 Cranioectodermal dysplasia
................................... ................................................................ ............................................................... ....................................................
Wdr34 Cilia/Hh P188R –6.337 Short-rib thoracic dysplasia
................................... ................................................................ ............................................................... ....................................................
Kif7 Cilia/Hh R833W –6.827 Joubert and acrocallosal
................................... ................................................................ ............................................................... ....................................................
Gli2 Hh P1086T –5.117 Culler-Jones
................................... ................................................................ ............................................................... ....................................................
Fat1 PCP S1717L
Y2462C
–5.858
–8.592
Facioscapulohumeral
dystrophy*
................................... ................................................................ ............................................................... ....................................................
Dchs1 PCP G2063D –6.45 Van Maldergem
................................... ................................................................ ............................................................... ....................................................
Dvl1 PCP P103L –8.23 Robinow
................................... ................................................................ ............................................................... ....................................................
*Based on phenotypic similarity to mutant mouse model.
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

Very similar numbers of variants and PROVEAN
score distributions were obtained for the other
homology-based (fig. S5) and transcriptome-based
annotations (fig. S3).
Enrichment for genes mutated in
skeletal ciliopathies
To identify proteins carrying function-altering
variants in the Galapagos cormorant, we applied
a stringent threshold to our four prediction data
sets: PROVEAN score less than –5, twice the thresh-
old for discovery of human disease variants (17,34)
(Fig. 2A). In our data set, variants with a PROVEAN
score less than –5 typically occur at residues that
have been perfectly conserved at least since mammals
and birds last shared a common ancestor (~30 0 Ma;
fig. S6). Consequently, changes in these residues
are likely to alter protein function or stability.
On the basis of theoretical and experimental
considerations (36), we hypothesized that flight-
lessness is likely to have a polygenic basis and
that the underlying variants would be enriched
in certain biological pathways. Consistent with this
hypothesis, gene enrichment analysis of function-
altering variants in the Galapagos cormorant
revealed that genes implicated in human develop-
mental disorders were significantly overrepresented
(17) (table S3A). Strikingly, 8 of the 19 significantly
enriched categories (Human Phenotype Ontology)
consisted of genes that affect limb development
when mutated, leading to disorders such as poly-
dactyly, syndactyly, and duplication of limb bones.
Control analyses showed no enrichment of these
categories in the flighted cormorants (17)(table
S3, B and C).
Many of the genes underlying the enrichment
for limb syndromes are those mutated in a family
of human disorders known as ciliopathies. For
instance, 17 of 25 genes (65%) in the “duplication
of hand bones”category and all 12 genes (100%)
in the “preaxial hand polydactyly”category are
mutated in human ciliopathies (table S4). More-
over, ciliopathy-associated genes were present in
all of the enriched categories (table S4). Ciliopathies
comprise a phenotypically diverse group of rare
genetic disorders that result from defects in the
formation or function of cilia (37).
Cilia are hairlike microtubule-based structures
that are nucleated by the basal body (centriole and
associated proteins) and project from the surface
of cells. Primary cilia are essential for mediating
Hedgehog (Hh) signaling in vertebrates, serving
as antennae for morphogens during development
(38). We confirmed by Sanger sequencing the
presence of predicted function-altering variants
in Ofd1,Evc,Talpid3,Dync2h1,Ift122,Wdr34,and
Kif7, all of which are necessary for the assembly
or functioning of the primary cilium and are mu-
tated in human ciliopathies, particularly those
affecting the skeleton (Table 1). We also found
a likely function-altering variant in Gli2,atran-
scription factor necessary for Hh signaling (39)
(Table 1). Humans affected by diverse skeletal
ciliopathies have small limbs and ribcages (37),
suggesting a parallel with the main features of
the Galapagos cormorant: small wings and a flat-
tened sternum. However, the consequences of
ciliopathies in humans are often more severe and
pleiotropic, likely as a consequence of the over-
representation of loss-of-function alleles in patients
(40). Interestingly, although ciliopathies do not
exclusively affect the forelimb in humans, differ-
encesbetweenforelimbsandhindlimbsarecom-
monly found among patients. For example, digital
abnormalitiesaffectthehandsmoreoftenthan
the feet among oral-facial-digital syndrome type I
(40,41)andEllis–van Creveld (42) patients, which
suggests that forelimbs and hindlimbs differ in
their intrinsic sensitivity to ciliary dysfunction.
The Galapagos cormorant ortholog of human
OFD1 (mutated in oral-facial-digital syndrome 1)
contains three predicted function-altering var-
iants with a PROVEAN score less than –5: Arg
325
→Cys (R325C), –6.913; Lys
517
→Thr (K517T),
–5.673;and Glu
889
→Gly (E889G), –5.068 (fig. S6,
AandB).Ofd1 knockout mice display polydactyly
and shortened long bones (43). Also, a function-
altering missense variant [Gln
691
→Leu (Q691L),
score –5.491] was found in IFT122, a component
oftheIFTcomplexthatcontrolsciliogenesisand
the ciliary localization of Shh pathway regulators
(44). Null Ift122 mutants show severe limb and
skeletal phenotypes in mice (45), and mutations
in Ift122 have been associated with Sensenbrenner
syndrome in humans, which is characterized by
craniofacial, ectodermal, and skeletal abnormal-
ities, including limb shortening (46). Strikingly,
the mutated Gln in IFT122 is virtually invariant
among eukaryotes as different as green algae,
C. elegans,Drosophila, and vertebrates (fig. S6C).
To directly test whether the nonsynonymous sub-
stitution in IFT122 affects protein function, we
generated a C. elegans knock-in strain using CRISPR/
Cas9 homology-directed genome editing (Fig. 3A).
The edited strain carries the Galapagos cormorant
missense variant at the corresponding orthologous
position in daf-10 [Gln
862
→Leu (Q862L)], the
ortholog of IFT122 in C. elegans (47).
In invertebrates, cilia do not mediate Shh sig-
naling but are necessary for detecting external sen-
sory inputs (48). The only ciliated cells in C. elegans
are sensory neurons, and mutations in cilia com-
ponents affect dispersal behavior, chemotaxis, and
dauer formation (49). We tested the bordering be-
havior of worms (accumulation of animals on the
thickest part of a bacterial lawn), which is known
to be mediated by ciliated neurons (50). We found
that daf-10(e1387) mutants carrying a premature
stop (Q892X) displayed an increase in bordering
behavior in a dispersal assay relative to the wild
type [73% for daf-10(e1387) versus 30% for N2; P=
0.019, ttest;Fig.3,B,C,andE].Twoindepen-
dently generated daf-10 Q862L knock-in lines
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 3of8
Fig. 2. Distribution of the effect of variants between P. auritus and P. harrisi.(A) We used PROVEAN to predict the effect on protein function of
23,402 variants contained in 12,442 orthologous pairs between P. auritus and P. harrisi; 4966 pairs contained no variants. The more negative the score,
the more likely the variant affects protein function. PROVEAN score thresholds used in this study are drawn as vertical dashed lines. Numbers of proteins
and variants found for each threshold are shown in the inset. (B) Density of PROVEAN scores for each class of variant.The same variants shown in (A) were
classified as single amino acid substitutions, deletions, and insertions. Numbers of variants in each class are indicated. a.u., arbitrary units.
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

phenocopied the effect of the premature stop
allele (70% for line 1 and 69% for line 2 versus
30% for N2; P= 0.018 and 0.016, respectively;
Fig. 3, B, D, and E; see movie S1). Daf-10(e1387)
ciliary neurons fail to incorporate fluorescent dyes,
like many other loss-of-function mutants in cilia
components (49). In contrast, the daf-10 Q862L
knock-in worms incorporated the DiO dye in the
same manner as wild-type worms, which suggests
that this allele is hypomorphic (fig. S7). Overall,
these results indicate that the IFT122 Q691L
missense variant present in the Galapagos cormo-
rantcanaffectproteinfunctioninvivoinC. elegans.
The planar cell polarity (PCP) pathway exhib-
its a genetic link to cilia (38,51–53). We found
function-altering variants in members of the PCP
pathway in P. harrisi: Fat1 atypical cadherin
(Fat1), Dachsous cadherin-related 1 (Dchs1), and
Disheveled-1 (Dvl1) (Table 1). The Galapagos cor-
morant FAT1 contains two function-altering var-
iants, Ser
1717
→Leu and Tyr
2462
→Cys (S1717L
and Y2462C; Table 1). The mutated Ser and Tyr
are conserved from zebrafish to humans (fig. S5D).
Fat1 knockout mice show very selective defects
in muscles of the upper body but not in posterior
muscles (54). In addition, Dvl1 is mutated in hu-
mans with Robinow syndrome, characterized by
limb shortening (55,56).
Sanger sequencing of 20 Galapagos cormorant
individuals from two different populations (Cabo
Hammond and Cañones Sur) (57)revealedonly
homozygous carriers for all of the variants in
Table 1, indicating that these variants are most
likely fixed in the Galapagos cormorant. In sum-
mary, we found an overrepresentation of pre-
dicted function-altering variants in genes that,
when mutated in humans and mice, cause skeletal
ciliopathies and bone growth defects.
CUX1 is mutated in P. harrisi
To identify the most likely function-altering var-
iants in P. harrisi,weappliedamorestringent
PROVEAN score threshold: –12.5 delta alignment
score, five times the threshold used for discovery
of human disease variants (34). This strategy nar-
rowed our search to 23 proteins (0.16% of anno-
tated proteins in P. harrisi)(tableS5).Wemanually
curated these 23 proteins and performed addi-
tional Sanger sequencing, reducing the list of pro-
teins with confirmed or putative variants to 12
(table S5) (17). These variants were exclusively
small deletions. Among these 12 proteins, two stood
out from their known role in development: LGALS-
3 and CUX1. LGALS-3 is affected by a 7–amino acid
deletion in P. harrisi (PROVEAN score, –26.319).
LGALS-3 (Galectin-3) is localized at the base of
the primary cilium and is necessary for correct
ciliogenesis in mice (58),butithasnotbeenim-
plicated in human ciliopathies. Moreover, LGALS-
3physicallyinteractswith SUFU, an important
regulator of mammalian Hh signaling (59), and
knockout mice show pleiotropic defects in chon-
drocyte differentiation (60).
In addition, we found a 4–amino acid deletion
(PROVEAN score, –15.704) in CUX1. CUX1 (cut-
like homeobox 1), also known as CDP, is a highly
conserved transcription factor with diverse roles
in development. CUX1 contains four DNA bind-
ing domains: three CUT domains (CR1 to CR3)
and one homeodomain (HD) (Fig. 4A) (61). The
full-length isoform, which contains four DNA bind-
ing domains (CR1-3HD), acts exclusively as a tran-
scriptional repressor and has rapid and unstable
DNA binding dynamics. In contrast, smaller iso-
forms such as CR2-3HD and CR3HD can act as
both repressors and activators of gene expression,
and show slow and stable DNA binding dynam-
ics in vitro (61).Althoughinsectandbirdwings
evolved independently, it is noteworthy that cut,
the Drosophila ortholog of Cux1, is necessary for
the proper development of wings and flight muscles
in flies (62). In chicken, Cux1 mRNA expression
in the limb at embryonic stage 23 is restricted to
the ectoderm bordering the apical ectodermal ridge
(AER) (63).TheAERisoneofthekeysignalingcen-
ters that drive limb development. At later stages,
Cux1 is expressed in the developing joints of both
chicken (64)andmouse(fig.S8)andisalsode-
tected in chondrocytes in developing bones of
mice (fig. S9). A function-altering variant in CUX1
is a strong candidate to contribute to loss of flight
in P. harrisi, because adenovirus-mediated over-
expression in the developing chicken wing of a
form of CUX1 missing the Cut2 DNA binding do-
main results in severe wing truncation (63,65).
These truncations most strongly affect distal skel-
etal elements (digits, radius, and ulna). As already
noted, in P. harrisi the radius and ulna are dis-
proportionately small relative to the humerus (7).
We Sanger-sequenced and confirmed the pre-
dicted Cux1 12–base pair (bp) deletion in P. harrisi.
We also confirmed that this variant is fixed in the
population and absent in the other cormorant
species (fig. S10A). The 12-bp deletion in Cux1 re-
moves four amino acids, Ala-Gly-Ser-Gln (AGSQ),
immediately adjacent to the C-terminal end of
the homeodomain (Fig. 4B). We refer to this var-
iant as CUX1-D4aa. Alignment of CUX1 orthologs
from available vertebrate genomes revealed that
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 4of8
Fig. 3. The Galapagos cormorant variant IFT122 Q691L affects ciliary function in vivo. (A) The
daf-10 gene (IFT122 ortholog) was targeted with CRISPR/Cas9 homology-mediated repair in
C. elegans to introduce a nonsynonymous substitution present exclusively in the Galapagos cormorant
(IFT122 Q691L). The resulting edited knock-in strain contains the daf-10 Q862L substitution and
10 synonymous substitutions (17). Edited strains were sequenced with Sanger sequencing to
confirm genotypes. gRNA, guide RNA. (Bto D) Representative bordering behavior of N2 wild-type
worms (B), daf-10(e1387) containing a premature stop codon Q892X (C), and daf-10 Q862L knock-in
strain (D). (E) Quantification of bordering behavior in N2, daf-10(e1387), and two independently
generated knock-in daf-10 Q862L strains (n=3,ttest). *P< 0.05; error bars denote SE.
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

the four missing residues are extremely con-
served among tetrapods (Fig. 4B). The deleted
Ser is phosphorylated in human cells (66), but the
consequences of this modification are unknown.
The Cux1 deletion does not include any of the
predicted residues responsible for DNA contact
and recognition (67), but given its close proximity
to the homeodomain, we decided to test whether
the DNA binding activity of CUX1 was affected.
We chose to express the CR3HD isoform because
Western blot analysis revealed that this was the
most abundant CUX1 isoform expressed in the
developing wing of mallard embryos (~50 kDa;
Fig. 4A). We performed electrophoretic mobility
shift assay (EMSA) with purified CR3HD CUX1-
Ancestral and CUX1-D4aa protein variants (fig.
S10B) as described (68) and found that DNA bind-
ing was not abolished in the deletion variant (fig.
S10C). CUX1 is able to both directly repress and
activate gene expression through its C-terminal
tail (69,70). We performed a luciferase reporter
assay (69,71) and found that both variants were
equally capable of repressing the expression of a
UAS/tk luciferase reporter (Fig. 4D). Thus, the
Galapagos cormorant CUX1-D4aa variant appears
tonotaffectDNAbindinginvitroortheC-terminal
repression activity in COS-7 cells.
CUX1 regulates the expression of cilia
and PCP genes
We hypothesized that the Cux1 deletion variant
is mechanistically related to the enrichment
of function-altering variants in ciliopathy-related
genes. This inference came from the fact that trans-
genic mice overexpressing the CR3HD-CUX1 iso-
form develop polycystic kidneys. Cilia in cystic
epithelial cells from these animals were twice
as long as the ones in control epithelial cells (72).
Furthermore, the CR2CR3HD-CUX1 isoform has
been shown to directly up-regulate the expression
of RPGRIPL1, also known as FTM, a component
of the cilia basal body that is involved in Shh
signaling and mutated in human ciliopathies (73).
Also, Cux1 knockout mice show deregulation of
SHH expression in hair follicles (74).
To test whether Cux1 could globally regulate
expression of cilia genes, we analyzed expression
array data from human-derived Hs578t cells stably
expressing a short hairpin RNA against Cux1,as
well as cells overexpressing the human CR2CR3HD-
CUX1 isoform (75). In concordance with the role
of Cux1 as a regulator of cell growth and pro-
liferation (76), genes significantly up- or down-
regulated in both conditions (P<0.05and>10%
change) were enriched for pathways such as “cell
cycle”and “mitotic G
1
-G
1
/S phases”(P=3.99×10
−5
and 0.016, respectively; table S6). We also found
enrichment for cilia-related categories such as
“assembly of the primary cilium”and “intraflagellar
transport”(P= 0.00012 and 0.0057, respectively;
table S6). These results suggest that cilia-related
genes are enriched among Cux1 targets.
To further test whether Cux1 can regulate cil-
iarygenesinanappropriatecellularcontext,we
generated ATDC5 stable lines expressing N-
terminal His-tagged versions of CR3HD CUX1-
Ancestral and CUX1-D4aa variants. ATDC5 is a
well-characterized mouse chondrogenic cell line
that largely recapitulates in vitro the differentia-
tion landmarks of chondrocytes (77). We performed
quantitative reverse transcription polymerase chain
reaction (RT-qPCR) on a selected number of genes
containing predicted strong function-altering var-
iants in P. harrisi (Table 1) and showing detect-
able levels of expression in ATDC5 cells. In addition,
we measured the expression of Ptch1,thereceptor
oftheHhpathway.Ourexperimentsindicatethat
the CUX1-Ancestral variant transcriptionally
up-regulated the expression of Ofd1 (factor of 1.7,
P=1.2×10
−6
;Fig.4C)andFat1 (factor of 1.8, P=
0.029; Fig. 4C) and down-regulated the expres-
sion of Ift122 (factor of 0.77, P=0.0025;Fig.4C)
and Ptch1 (factor of 0.53, P=0.014;Fig.4C)relative
to the control line. In contrast, neither Dync2h1
nor Wdr34 expression levels were changed by
CUX1-Ancestral overexpression (Fig. 4C). These
results suggest that cilia- and Hh-related genes
arelikelytranscriptionaltargetsofCUX1in
chondrocytes.
Impaired transcriptional activity of the
Galapagos cormorant CUX1
The Galapagos cormorant CUX1 showed impaired
transcriptional activity relative to the ancestral var-
iant. Ofd1 was significantly up-regulated in CUX1-
D4aa cells relative to control cells [factor of 1.2, P=
0.021, analysis of variance (ANOVA) and Tukey
honest significant difference (HSD) test, Fig. 4C];
however, Ofd1 up-regulation was significantly
reduced in CUX1-D4aa cells relative to CUX1-
Ancestral cells (factor of 1.2 versus 1.7, P=6×10
−5
;
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 5of8
Fig. 4. The Galapagos cormorant Cu x1 is a transcriptional activation hypomorph. (A) Western
blot showing the expression of CUX1 isoforms in the developing wing of a mallard embryo (22 days).
The most abundant band corresponds to the predicted size of the CR3HD CUX1 isoform. (B)Protein
alignment showing the deleted AGSQ residues in the Galapagos cormorant CUX1 and their high degree
of conservation among vertebrates. (C) Differential up-regulation of genes by CUX1-Ancestral and
CUX1-D4aa variants in ATDC5 cells. Pooled stable lines carrying CR3HD CUX1-Ancestral or CUX1-D4aa
variants were generated by lentiviral transduction and puromycin selection. Control cells were
transduced with an empty vector. Gene expression levels were measured by RT-qPCR (n= 5 biological
replicates, each comprising three technical replicates). (D) Luciferase-based assay to test the
repression activity of CUX1 C-terminal domain lacking CR3 and HD domains. GAL4 DNA binding
domain was fused to CUX1-Ancestral or CUX1-D4aa variants. Both constructs equally repressed a
promoter containing UAS binding sites in COS-7 cells (n= 3 biological replicates, each comprising three
technical replicates). In (C) and (D), error bars denote SE. *P<0.05,**P<0.01,***P< 0.001 (ANOVA
and Tukey HSD); black and red asterisks respectively denote significant difference of CUX1 variant
versus control and CUX1-Ancestral variant versus CUX1-D4aa variant.
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

Fig. 4C). Similarly, although Fat1 was significantly
up-regulated in CUX1-Ancestral cells relative to
control cells (factor of 1.8, P=0.029),Fat1 ex-
pression levels in CUX1-D4aa cells were not sig-
nificantly different from control lines (factor of 1.2,
P= 0.57; Fig. 4C). The difference between Fat1
up-regulation in CUX1-D4aa and CUX1-D4aa cells
was not significant (factor of 1.8 versus 1.3, P=
0.17; Fig. 4C). In contrast, CUX1-D4aa cells signif-
icantly repressed both Ift122 (factor of 0.78, P=
0.0026) and Ptch1 (factor of 0.56, P=0.024),and
there were no significant differences between
CUX1-Ancestral and CUX1-D4aa cells (factor of
0.78 versus 0.78 for Ift122,P= 0.99; factor of 0.53
versus 0.56 for Ptch1,P= 0.96; Fig. 4C). These
results suggest that the 4–amino acid deletion in
the Galapagos cormorant CUX1 affects its ability
to activate but not to repress gene expression;
they are also consistent with our luciferase reporter
assays, which showed no effect on repression (Fig.
4D). It is notable that both the transcriptional
activator (Cux1) and its target genes (Ofd1 and
Fat1) exhibit function-altering variants in the
Galapagos cormorant.
CR3HD-CUX1 promotes chondrogenesis
Chondrocytes are the main engine of bone growth.
The growth of skeletal elements depends on the
precise regulation of chondrocyte proliferation and
hypertrophy. Mutations that affect cilia result in
premature arrest of bone growth due to defects in
Indian Hedgehog (IHH) signaling in chondrocytes
(78). To test the role of CR3HD-CUX1 in chondro-
genesis, we differentiated control, CUX1-Ancestral,
and CUX1-D4aa ATDC5 cell lines and quantified
the expression of Ihh and Sox9, two well-established
markers of chondrocyte differentiation in vitro
and in vivo (78). Overexpression of both CR3HD
CUX1-Ancestral and CUX1-D4aa variants promoted
chondrogenic differentiation of ATDC5 cells after
7 and 12 days of differentiation (Fig. 5A). How-
ever, the CUX1-D4aa variant was not as efficient
as the ancestral variant, showing significant dif-
ferences from CUX1-Ancestral in Ihh expression
after 7 days of differentiation (~50% decrease, P=
5.9×10
−4
,ANOVAandTukeyHSDtest;Fig.5A)
and in Sox9 expression after 12 days (~15% de-
crease, P=1.6×10
−2
; Fig. 5A). These results
suggest that the Galapagos cormorant CUX1 is
probably not as effective as the ortholog from
its flighted relatives in promoting chondrogenic
differentiation, and that mutations in Cux1 may
affect the dynamics of chondrogenesis. This obser-
vation is further supported by findings that CUX1
is expressed in the hypertrophic chondrocytes of
developing bones in mice, and that the bones of
Cux1 mutant mice are thin and flaky (79).
Possible evolutionary scenarios
Loss of flight has traditionally been attributed to
relaxed selection. In this scenario, the first cor-
morants that inhabited the Galapagos Islands
found a unique environment that lacked preda-
tors and provided food year-round, drastically
reducing the need to migrate. However, we found
no evidence for pseudogenization of developmen-
tal genes in P. harrisi (17) (tables S10 and S11). On
the other hand, loss of flight in the Galapagos
cormorant is thought to confer an advantage for
diving by decreasing buoyancy via shorter wings
and by indirectly allowing increased oxygen stor-
age via larger body size (80). This advantage could
make flightlessness a target of positive selection.
We evaluated whether any of our candidate
genes (Table 1) showed signatures of positive
selectionintheGalapagoscormorantlineageby
estimating the ratio of nonsynonymous to syn-
onymous substitutions (w=dN/dS). This is a very
stringent test of selection because it assumes that
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 6of8
Fig. 5. CR3HD CUX1 promotes chondrogenesis. (A) ATDC5 control cells and cells carrying CR3HD
CUX1-Ancestral or CUX1-D4aa variants were differentiated into chondrocytes. Gene expression levels
were measured by RT-qPCR (n= 4 biological replicates, each comprising three technical replicates)
after 7 and 12 days. Error bars denote SE. *P<0.05,**P<0.01,***P< 0.001 (ANOVA and Tukey
HSD); black and red asterisks respectively denote significant difference of CUX1 variant versus control
and CUX1-Ancestral variant versus CUX1-D4aa variant. (B) Proposed mechanism for the reduction
of wing size in P. harrisi. Left: Normal functioning of IHH signaling pathway in vertebrates. CUX1
regulates the expression of cilia-related genes such as Ofd1 and promotes chondrogenesis. Right:
State of the IHH pathway in P. harrisi. Proteins in red have predicted function-altering variants in
P. harrisi. We propose that these variants would affect both cilia formation and functioning, leading
to a reduction in IHH pathway activity. As a result, the pool of proliferating chondrocytes would
decrease in wing bones and the number of hypertrophic chondrocytes would increase, resulting in
impaired bone growth.
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

all sites in a protein are evolving under the same
selective pressure, a condition rarely met in highly
conserved regulatory genes (81). We found that 3
of 11 tested genes showed signs of positive selec-
tion (w>1)intheGalapagoscormorantlineage
compared to a background phylogeny of 35 taxa
(Ofd1 w=1.92,Evc w=1.93,Gli2 w=1.10;tableS8).
Oneofthesethreegenes,Gli2,showedastatis-
tically significant difference [w=1.10(Galapagos
branch) versus w= 0.11 (background branch), P=
0.0024; table S8]. In contrast, Gli2 showed no
sign of selection in the sister group of P. harrisi
(P. auritus and P. brasilianus)[w= 0.0001 (sister
branch) versus w= 0.11 (background branch), P=
0.46]. As a control, we also analyzed Gli3,the
partially redundant paralog of Gli2,whichalsomedi-
ates Hh signaling but has no predicted function-
altering variants in P. harrisi, and found no evidence
for positive selection [w=0.04(Galapagosbranch)
versus w= 0.15 (background branch), P=0.11;
table S8]. These results suggest that selection
toward flightlessness may be partially respon-
sible for the phenotype of P. harrisi.
Discussion
The study of evolution of flightlessness in the
Rallidaefamilyledtothehypothesisthatflight-
lessness could be a fast-evolving heterochronic con-
dition (10,82). Heterochrony, the relative change
in the rate or timing of developmental events
among species, is thought to be an important fac-
tor contributing to macroevolutionary change (83).
Yet virtually nothing is known about its genetic
and molecular mechanisms.
Diverse myological, osteological, and develop-
mental observations suggest that flightlessness
in the Galapagos cormorant is caused by the re-
tention into adulthood of juvenile characteristics
affecting pectoral and forelimb development (a
class of heterochrony known as paedomorpho-
sis) (7). Here, we propose a genetic and molecular
model that may explain this heterochronic con-
dition, where the perturbations of cilia/Ihh sig-
naling may be responsible for the reduction in
growth of both keel and wings in the Galapagos
cormorant (Fig. 5B). However, we cannot rule
out a role of Cux1 in the AER. Of special interest
is the gene Fat1,atargetofCux1 (Fig. 4D), which
contains two putative function-altering variants
(Table 1). Fat1
−/−
mousemutantsareviableand
show selective defects in facial, pectoral, and
shoulder muscles but not in hindlimb muscles
(54). Thus, variants in Fat1 could explain the
underdeveloped pectoral muscle s of P. harrisi.
Although we have identified multiple variants
that likely contribute to the flightless phenotype
of P. harrisi,wecannotexcludethepossibility
that other genes and pathways may contribute
to the phenotype, nor the contribution of noncoding
regulatory variants (17). Further characterization
of the individual and joint contributions of the
variants found in this study will help us to recon-
struct the chain of events leading to flightless-
ness and to genetically dissect macroevolutionary
change. We hypothesize that mutations in cilia or
functionally related genes could be responsible
for limb and other skeletal heterochronic trans-
formations in birds and diverse organisms, in-
cluding humans.
REFERENCES AND NOTES
1. D. A. Roff, The evolution of flightlessness: Is history important?
Evol. Ecol. 8, 639–657 (1994). doi: 10.1007/BF01237847
2. C. Darwin, On the Origin of Species by Means of Natural
Selection, or the Preservation of Favoured Races in the Struggle
for Life (John Murray, London, 1859).
3. J. Harshman et al., Phylogenomic evidence for multiple losses
of flight in ratite birds. Proc. Natl. Acad. Sci. U.S.A. 105,
13462–13467 (2008). doi: 10.1073/pnas.0803242105;
pmid: 18765814
4. A. J. Baker, O. Haddrath, J. D. McPherson, A. Cloutier, Genomic
support for a moa-tinamou clade and adaptive morphological
convergence in flightless ratites. Mol. Biol. Evol. 31,1686–1696
(2014). doi: 10.1093/molbev/msu153;pmid:24825849
5. K. J. Mitchell et al., Ancient DNA reveals elephant birds and kiwi
are sister taxa and clarifies ratite bird evolution. Science 344,
898–900 (2014). doi: 10.1126/science.1251981;
pmid: 24855267
6. N.A.Wright,D.W.Steadman,C.C.Witt,Predictable
evolution toward flightlessness in volant island birds.
Proc. Natl. Acad. Sci. U.S.A. 113,4765–4770 (2016).
doi: 10.1073/pnas.1522931113;pmid:27071105
7. B. C. Livezey, Flightlessness in the Galápagos cormorant
(Compsohalieus [Nannopterum] harrisi): Heterochrony,
gigantism, and specialization. Zool. J. Linn. Soc. 105, 155–224
(1992). doi: 10.1111/j.1096-3642.1992.tb01229.x
8. M. A. G. de Bakker et al., Digit loss in archosaur evolution and
the interplay between selection and constraints. Nature 500,
445–448 (2013). doi: 10.1073/pnas.1522931113;
pmid: 27071105
9. The onset times of several homologous developmental events,
such as the appearance of immature feathers, dropping of egg
tooth, and independence from parental care, occur progres-
sively later in the Galapagos cormorant chick than in P. auritus.
10. A. Feduccia, The Origin and Evolution of Birds (Yale Univ. Press,
ed. 2, 1999).
11. K. E. Slack et al., Early penguin fossils, plus mitochondrial
genomes, calibrate avian evolution. Mol. Biol. Evol. 23,
1144–1155 (2006). doi: 10.1093/molbev/msj124;
pmid: 16533822
12. M. Kennedy, C. A. Valle, H. G. Spencer, The phylogenetic
position of the Galápagos cormorant. Mol. Phylogenet. Evol. 53,
94–98 (2009). doi: 10.1016/j.ympev.2009.06.002;
pmid: 19523526
13. M. Kennedy, H. G. Spencer, Classification of the cormorants of
the world. Mol. Phylogenet. Evol. 79, 249–257 (2014).
doi: 10.1016/j.ympev.2014.06.020; pmid: 24994028
14. A recent phylogeny of the Phalacrocoracidae suggested
classifying P. harrisi,P. auritus, and P. brasilianus as members
of the monophyletic genera Nannopterum.
15. R. Luo et al., SOAPdenovo2: An empirically improved memory-
efficient short-read de novo assembler. Gigascience 1,18
(2012). doi: 10.1186/2047-217X-1-18; pmid: 23587118
16. G. Parra, K. Bradnam, Z. Ning, T. Keane, I. Korf, Assessing the
gene space in draft genomes. Nucleic Acids Res. 37, 289–297
(2009). doi: 10.1093/nar/gkn916; pmid: 19042974
17. See supplementary materials.
18. International Chicken Genome Sequencing Consortium,
Sequence and comparative analysis of the chicken genome
provide unique perspectives on vertebrate evolution. Nature
432, 695–716 (2004). doi: 10.1038/nature03154;
pmid: 15592404
19. W. C. Warren et al., The genome of a songbird. Nature 464,
757–762 (2010). doi: 10.1038/nature08819; pmid: 20360741
20. G. Ganapathy et al., High-coverage sequencing and annotated
assemblies of the budgerigar genome. Gigascience 3,11
(2014). doi: 10.1186/2047-217X-3-11; pmid: 25061512
21. D. J. Geist, H. Snell, H. Snell, C. Goddard, M. D. Kurz, in The
Galápagos: A Natural Laboratory for the Earth Sciences,
K. S. Harpp, E. Mittelstaedt, N. D’Ozouville, D. W. Graham, Eds.
(Wiley, 2014), pp. 145–166.
22. G. Wang, X. Li, Population dynamics and recovery of
endangered crested ibis (Nipponia nippon) in central China.
Waterbirds 31, 489–494 (2008). doi: 10.1675/
1524-4695-31.3.489
23. C. A. Valle, M. C. M. Coulter, Present status of the flightless
cormorant, Galapagos penguin and greater flamingo
populations in the Galapagos Islands, Ecuador, after the 1982-
83 El Nino. Condor 89, 276–281 (1987). doi: 10.2307/1368480
24. J. Alföldi, K. Lindblad-Toh, Comparative genomics as a tool to
understand evolution and disease. Genome Res. 23,
1063–1068 (2013). doi: 10.1101/gr.157503.113;
pmid: 23817047
25. R. Jelier, J. I. Semple, R. Garcia-Verdugo, B. Lehner, Predicting
phenotypic variation in yeast from individual genome
sequences. Nat. Genet. 43, 1270–1274 (2011). doi: 10.1038/
ng.1007; pmid: 22081227
26. M. D. Shapiro et al., Genetic and developmental basis of
evolutionary pelvic reduction in threespine sticklebacks. Nature
428, 717–723 (2004). doi: 10.1038/nature02415;
pmid: 15085123
27. D. L. Stern, V. Orgogozo, The loci of evolution: How predictable
is genetic evolution? Evolution 62, 2155–2177 (2008).
doi: 10.1111/j.1558-5646.2008.00450.x; pmid: 18616572
28. K. S. Pollard et al., An RNA gene expressed during cortical
development evolved rapidly in humans. Nature 443, 167–172
(2006). doi: 10.1038/nature05113; pmid: 16915236
29. B. M. Booker et al., Bat accelerated regions identify a bat
forelimb specific enhancer in the HoxD locus. PLOS Genet. 12,
e1005738 (2016). doi: 10.1371/journal.pgen.1005738;
pmid: 27019019
30. S. Prabhakar et al., Human-specific gain of function in a
developmental enhancer. Science 321, 1346–1350 (2008).
doi: 10.1126/science.1159974; pmid: 18772437
31. T. M. Frayling et al., A common variant in the FTO gene is
associated with body mass index and predisposes to childhood
and adult obesity. Science 316, 889–894 (2007).
doi: 10.1126/science.1141634; pmid: 17434869
32. A. Visel et al., ChIP-seq accurately predicts tissue-specific
activity of enhancers. Nature 457, 854–858 (2009).
doi: 10.1038/nature07730; pmid: 19212405
33. J. Cotney et al., Chromatin state signatures associated with
tissue-specific gene expression and enhancer activity in the
embryonic limb. Genome Res. 22, 1069–1080 (2012).
doi: 10.1101/gr.129817.111; pmid: 22421546
34. Y. Choi, G. E. Sims, S. Murphy, J. R. Miller, A. P. Chan,
Predicting the functional effect of amino acid substitutions and
indels. PLOS ONE 7, e46688 (2012). doi: 10.1371/journal.
pone.0046688; pmid: 23056405
35. L. Rockah-Shmuel, Á. Tóth-Petróczy, D. S. Tawfik, Systematic
mapping of protein mutational space by prolonged drift reveals
the deleterious effects of seemingly neutral mutations. PLOS
Comput. Biol. 11, e1004421 (2015). doi: 10.1371/journal.
pcbi.1004421; pmid: 26274323
36. D. L. Stern, Perspective: Evolutionary developmental biology
and the problem of variation. Evolution 54, 1079–1091 (2000).
doi: 10.1111/j.0014-3820.2000.tb00544.x
37. A. M. Waters, P. L. Beales, Ciliopathies: An expanding disease
spectrum. Pediatr. Nephrol. 26, 1039–1056 (2011).
doi: 10.1007/s00467-010-1731-7; pmid: 21210154
38. S. C. Goetz, K. V. Anderson, The primary cilium: A signalling
centre during vertebrate development. Nat. Rev. Genet. 11,
331–344 (2010). doi: 10.1038/nrg2774; pmid: 20395968
39. C. J. Haycraft et al., Gli2 and Gli3 localize to cilia and require
the intraflagellar transport protein polaris for processing and
function. PLOS Genet. 1, e53 (2005). doi: 10.1371/journal.
pgen.0010053; pmid: 16254602
40. I. J. Bisschoff et al., Novel mutations including deletions of the
entire OFD1 gene in 30 families with type 1 orofaciodigital
syndrome: A study of the extensive clinical variability. Hum.
Mutat. 34, 237–247 (2013). doi: 10.1002/humu.22224;
pmid: 23033313
41. K. L. M. Coene et al., OFD1 is mutated in X-linked Joubert
syndrome and interacts with LCA5-encoded lebercilin.
Am.J.Hum.Genet.85, 465–481 (2009). doi: 10.1016/
j.ajhg.2009.09.002;pmid:19800048
42. G. Baujat, M. Le Merrer, Ellis-van Creveld syndrome. Orphanet
J. Rare Dis. 2, 27 (2007). doi: 10.1186/1750-1172-2-27;
pmid: 17547743
43. S. Bimonte et al., Ofd1 is required in limb bud patterning
and endochondral bone development. Dev. Biol. 349,
179–191 (2011). doi: 10.1016/j.ydbio.2010.09.020
pmid: 20920500
44. P. J. R. Ocbina, J. T. Eggenschwiler, I. Moskowitz,
K. V. Anderson, Complex interactions between genes
controlling trafficking in primary cilia. Nat. Genet. 43, 547–553
(2011). doi: 10.1038/ng.832; pmid: 21552265
45. J. Qin, Y. Lin, R. X. Norman, H. W. Ko, J. T. Eggenschwiler,
Intraflagellar transport protein 122 antagonizes Sonic
Hedgehog signaling and controls ciliary localization of pathway
components. Proc. Natl. Acad. Sci. U.S.A. 108, 1456–1461
(2011). doi: 10.1073/pnas.1011410108; pmid: 21209331
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 7of8
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

46. J. Walczak-Sztulpa et al., Cranioectodermal dysplasia,
Sensenbrenner syndrome, is a ciliopathy caused by mutations
in the IFT122 gene. Am. J. Hum. Genet. 86, 949–956 (2010).
doi: 10.1016/j.ajhg.2010.04.012; pmid: 20493458
47. L. R. Bell, S. Stone, J. Yochem, J. E. Shaw, R. K. Herman,
The molecular identities of the Caenorhabditis elegans
intraflagellar transport genes dyf-6, daf-10 and osm-1. Genetics
173, 1275–1286 (2006). doi: 10.1534/genetics.106.056721;
pmid: 16648645
48. V. Singla, J. F. Reiter, The primary cilium as the cell’s antenna:
Signaling at a sensory organelle. Science 313, 629–633
(2006). doi: 10.1126/science.1124534; pmid: 16888132
49. L. A. Perkins, E. M. Hedgecock, J. N. Thomson, J. G. Culotti,
Mutant sensory cilia in the nematode Caenorhabditis elegans.
Dev. Biol. 117, 456–487 (1986). doi: 10.1016/0012-1606(86)
90314-3; pmid: 2428682
50. E. Z. Macosko et al., A hub-and-spoke circuit drives pheromone
attraction and social behaviour in C. elegans.Nature 458,
1171–1175 (2009). doi: 10.1038/nature07886; pmid: 19349961
51. T. J. Park, S. L. Haigo, J. B. Wallingford, Ciliogenesis defects in
embryos lacking inturned or fuzzy function are associated
with failure of planar cell polarity and Hedgehog signaling.
Nat. Genet. 38,303–311 (2006). doi: 10.1038/ng1753;
pmid: 16493421
52. R. S. Gray et al., The planar cell polarity effector Fuz is
essential for targeted membrane trafficking, ciliogenesis and
mouse embryonic development. Nat. Cell Biol. 11, 1225–1232
(2009). doi: 10.1038/ncb1966; pmid: 19767740
53. H. Zeng, A. N. Hoover, A. Liu, PCP effector gene Inturned is an
important regulator of cilia formation and embryonic
development in mammals. Dev. Biol. 339, 418–428 (2010).
doi: 10.1016/j.ydbio.2010.01.003; pmid: 20067783
54. N. Caruso et al., Deregulation of the protocadherin gene FAT1
alters muscle shapes: Implications for the pathogenesis of
facioscapulohumeral dystrophy. PLOS Genet. 9, e1003550
(2013). doi: 10.1371/journal.pgen.1003550; pmid: 23785297
55. J. White et al., DVL1 frameshift mutations clustering in the
penultimate exon cause autosomal-dominant Robinow
syndrome. Am. J. Hum. Genet. 96, 612–622 (2015).
doi: 10.1016/j.ajhg.2015.02.015; pmid: 25817016
56. K. J. Bunn et al., Mutations in DVL1 cause an osteosclerotic
form of Robinow syndrome. Am. J. Hum. Genet. 96, 623–630
(2015). doi: 10.1016/j.ajhg.2015.02.010; pmid: 25817014
57. C. V. Duffie, T. C. Glenn, F. H. Vargas, P. G. Parker, Genetic
structure within and between island populations of the
flightless cormorant (Phalacrocorax harrisi). Mol. Ecol. 18,
2103–2111 (2009). doi: 10.1111/j.1365-294X.2009.04179.x;
pmid: 19635072
58. A. Koch, F. Poirier, R. Jacob, D. Delacour, Galectin-3, a novel
centrosome-associated protein, required for epithelial
morphogenesis. Mol. Biol. Cell 21, 219–231 (2010).
doi: 10.1091/mbc.E09-03-0193; pmid: 19923323
59. M. Paces-Fessy, D. Boucher, E. Petit, S. Paute-Briand,
M.-F. Blanchet-Tournier, The negative regulator of Gli,
Suppressor of fused (Sufu), interacts with SAP18, Galectin3
and other nuclear proteins. Biochem. J. 378, 353–362 (2004).
doi: 10.1042/bj20030786; pmid: 14611647
60. C. Colnot, S. S. Sidhu, N. Balmain, F. Poirier, Uncoupling of
chondrocyte death and vascular invasion in mouse galectin 3
null mutant bones. Dev. Biol. 229, 203–214 (2001).
doi: 10.1006/dbio.2000.9933; pmid: 11133164
61. L. Sansregret, A. Nepveu, The multiple roles of CUX1: Insights
from mouse models and cell-based assays. Gene 412,84–94
(2008). doi: 10.1016/j.gene.2008.01.017; pmid: 18313863
62. V. Sudarsan, S. Anant, P. Guptan, K. VijayRaghavan, H. Skaer,
Myoblast diversification and ectodermal signaling in
Drosophila.Dev. Cell 1, 829–839 (2001). doi: 10.1016/
S1534-5807(01)00089-2; pmid: 11740944
63. A. T. Tavares, T. Tsukui, J. C. Izpisúa Belmonte, Evidence that
members of the Cut/Cux/CDP family may be involved in
AER positioning and polarizing activity during chick limb
development. Development 127, 5133–5144 (2000).
pmid: 11060239
64. G. Lizarraga, A. Lichtler, W. B. Upholt, R. A. Kosher,
Studies on the role of Cux1 in regulation of the onset of
joint formation in the developing limb. Dev. Biol. 243,
44–54 (2002). doi: 10.1006/dbio.2001.0559;
pmid: 11846476
65. The chicken CUX1-WT variant studied by Tavares et al.
was inadvertently missing the Cut2 DNA binding domain,
as revealed by the chicken genome generated subsequently
to their study (Ensembl ID: ENSGALP00000002788).
This isoform is likely hypomorphic, and its overexpression
may exert a dominant negative effect.
66. S. Matsuoka et al., ATM and ATR substrate analysis reveals
extensive protein networks responsive to DNA damage.
Science 316, 1160–1166 (2007). doi: 10.1126/science.1140321;
pmid: 17525332
67. D. Iyaguchi, M. Yao, N. Watanabe, J. Nishihira, I. Tanaka, DNA
recognition mechanism of the ONECUT homeodomain of
transcription factor HNF-6. Structure 15,75–83 (2007).
doi: 10.1016/j.str.2006.11.004; pmid: 17223534
68. N. S. Moon, G. Bérubé, A. Nepveu, CCAAT displacement
activity involves CUT repeats 1 and 2, not the CUT
homeodomain. J. Biol. Chem. 275, 31325–31334 (2000).
doi: 10.1074/jbc.M002912200; pmid: 10864926
69. F. Mailly et al., The human cut homeodomain protein can
repress gene expression by two distinct mechanisms: Active
repression and competition for binding site occupancy. Mol.
Cell. Biol. 16, 5346–5357 (1996). doi: 10.1128/MCB.16.10.5346;
pmid: 8816446
70. M. Truscott et al., Carboxyl-terminal proteolytic
processing of CUX1 by a caspase enables transcriptional
activation in proliferating cells. J. Biol. Chem. 282,
30216–30226 (2007). doi: 10.1074/jbc.M702328200;
pmid: 17681953
71. H. Nishio, M. J. Walsh, CCAAT displacement protein/cut
homolog recruits G9a histone lysine methyltransferase to
repress transcription. Proc. Natl. Acad. Sci. U.S.A. 101,
11257–11262 (2004). doi: 10.1073/pnas.0401343101;
pmid: 15269344
72. C. Cadieux et al., Polycystic kidneys caused by sustained
expression of Cux1 isoform p75. J. Biol. Chem. 283,
13817–13824 (2008). doi: 10.1074/jbc.M709332200;
pmid: 18356167
73. G. Stratigopoulos, C. A. LeDuc, M. L. Cremona, W. K. Chung,
R. L. Leibel, Cut-like homeobox 1 (CUX1) regulates expression
of the fat mass and obesity-associated and retinitis
pigmentosa GTPase regulator-interacting protein-1-like
(RPGRIP1L) genes and coordinates leptin receptor signaling.
J. Biol. Chem. 286, 2155–2170 (2011). doi: 10.1074/
jbc.M110.188482; pmid: 21037323
74. T. Ellis et al., The transcriptional repressor CDP (Cutl1) is
essential for epithelial cell differentiation of the lung and the
hair follicle. Genes Dev. 15, 2307–2319 (2001). doi: 10.1101/
gad.200101; pmid: 11544187
75. C. Vadnais et al., Long-range transcriptional regulation by the
p110 CUX1 homeodomain protein on the ENCODE array. BMC
Genomics 14, 258 (2013). doi: 10.1186/1471-2164-14-258;
pmid: 23590133
76. Z. M. Ramdzan, A. Nepveu, CUX1, a haploinsufficient tumour
suppressor gene overexpressed in advanced cancers. Nat. Rev.
Cancer 14, 673–682 (2014). doi: 10.1038/nrc3805;
pmid: 25190083
77. C. Shukunami et al., Chondrogenic differentiation of clonal
mouse embryonic cell line ATDC5 in vitro: Differentiation-
dependent gene expression of parathyroid hormone
(PTH)/PTH-related peptide receptor. J. Cell Biol.
133,457–468 (1996). doi: 10.1083/jcb.133.2.457;
pmid: 8609176
78. H. M. Kronenberg, Developmental regulation of the growth
plate. Nature 423, 332–336 (2003). doi: 10.1038/nature01657;
pmid: 12748651
79. A. M. Sinclair et al., Lymphoid apoptosis and myeloid
hyperplasia in CCAAT displacement protein mutant mice.
Blood 98, 3658–3667 (2001). doi: 10.1182/blood.V98.13.3658;
pmid: 11739170
80. Y. Y. Watanabe, A. Takahashi, K. Sato, M. Viviant, C.-A.
Bost, Poor flight performance in deep-diving cormorants.
J. Exp. Biol. 214,412–421 (2011). doi: 10.1242/jeb.050161;
pmid: 21228200
81. Z. Yang, Likelihood ratio tests for detecting positive selection
and application to primate lysozyme evolution. Mol. Biol. Evol.
15,568–573 (1998). doi: 10.1093/oxfordjournals.molbev.
a025957
82. S. L. Olson, Evolution of the rails of the South Atlantic islands
(Aves: Rallidae). Smithson. Contrib. Zool. 152,1–53 (1973).
doi: 10.5479/si.00810282.152
83. S. J. Gould, Ontogeny and Phylogeny (Belknap/Harvard Univ.
Press, 1977).
ACKNO WLEDGM ENTS
Supported by the Jane Coffin Childs Memorial Fund for Medical
Research (A.B.), the Howard Hughes Medical Institute (L.K.), the
U.S. Geological Survey through the Wildlife Program of the
Ecosystems Mission Area (A.M.R.), a Gruss-Lipper postdoctoral
fellowship from the EGL Charitable Foundation (E.B.-D.),
FONDECYT grant 11130305 (C.V.), and the National Institute of
Arthritis and Musculoskeletal and Skin Diseases (K.L.). Field
collection permits and export permits were by the Galapagos
National Park, facilitated by the Charles Darwin Foundation. Any
use of trade names is for descriptive purposes only and does not
imply endorsement by the U.S. government. We thank C. Koo and
M. Arbing at the UCLA-DOE Protein Expression Laboratory Core
Facility for protein purification; S. Feng at the Broad Stem Cell
Research Center High Throughput Sequencing Core for assistance;
C. Cicero for granting access to cormorant specimens (134079
and 151575) at the Museum of Vertebrate Zoology at Berkeley;
C. Valle (Universidad San Francisco de Quito) for advice; and
K. Garrett (Natural History Museum of Los Angeles), K. Burns
(San Diego Natural History Museum and San Diego State
University Museum of Biodiversity), and R. Duerr (International
Bird Rescue) for providing samples used in preliminary stages of
this study. Author contributions: A.B. and L.K. conceived the study;
A.B. coordinated the collection of samples, prepared libraries,
assembled and annotated genomes, and performed analyses
and experiments; E.B.-D. and A.B. performed the accelerated
evolution analysis; P.C.W., A.M.R., C.V., and P.G.P. provided DNA or
tissue samples; W.W. and A.B. carried out ATDC5 cell line
experiments supervised by K.L.; A.B. and L.K. wrote the
manuscript; and all authors discussed and agreed on the final
version of the manuscript. All sequencing data from this study is
available through the NCBI Sequence Read Archive under
Bioproject accession number PRJNA327123. Alignments used for
phylogenetic analysis and selection test are available at DRYAD
doi:10.5061/dryad.8m2t5. The authors declare no competing
financial interests.
SUPPLEMENTARY MATERIALS
www.sciencemag.org/content/356/6341/eaal3345/suppl/DC1
Materials and Methods
Figs. S1 to S10
Tables S1 to S13
Movie S1
References (84–109)
4 November 2016; accepted 13 April 2017
10.1126/science.aal3345
Burga et al., Science 356, eaal3345 (2017) 2 June 2017 8of8
RESEARCH |RESEARCH ARTICLE
on June 13, 2017http://science.sciencemag.org/Downloaded from

(6341), . [doi: 10.1126/science.aal3345]356Science
Parker and Leonid Kruglyak (June 1, 2017)
Andrew M. Ramey, Claudio Verdugo, Karen Lyons, Patricia G.
Alejandro Burga, Weiguang Wang, Eyal Ben-David, Paul C. Wolf,
Galapagos cormorant
A genetic signature of the evolution of loss of flight in the
Editor's Summary
, this issue p. eaal3345; see also p. 904Science
cormorant's loss of flight.
function of the genes may be responsible for skeletal changes associated with the Galapagos
involved in primary ciliogenesis. Functional analyses of these variants suggest that the impaired
flightless Galapagos cormorant (see the Perspective by Cooper). They identified variants in genes
. sequenced genomes from three cormorant species and compared them with that of theet alBurga
evolutionary history. However, the genetic changes that ground avian species are not well understood.
Although rare among existing birds, the loss of flight appears to have occurred multiple times in
Loss of flight in the Galapagos cormorant
This copy is for your personal, non-commercial use only.
Article Tools
http://science.sciencemag.org/content/356/6341/eaal3345
article tools:
Visit the online version of this article to access the personalization and
Permissions http://www.sciencemag.org/about/permissions.dtl
Obtain information about reproducing this article:
is a registered trademark of AAAS. ScienceAdvancement of Science; all rights reserved. The title
Avenue NW, Washington, DC 20005. Copyright 2016 by the American Association for the
in December, by the American Association for the Advancement of Science, 1200 New York
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last weekScience
on June 13, 2017http://science.sciencemag.org/Downloaded from