New evidence for genome-wide duplications at the origin of vertebrates using an amphioxus gene set and completed animal genomes.

Abstract

The 2R hypothesis predicting two genome duplications at the origin of vertebrates is highly controversial. Studies published so far include limited sequence data from organisms close to the hypothesized genome duplications. Through the comparison of a gene catalog from amphioxus, the closest living invertebrate relative of vertebrates, to 3453 single-copy genes orthologous between Caenorhabditis elegans (C), Drosophila melanogaster (D), and Saccharomyces cerevisiae (Y), and to Ciona intestinalis ESTs, mouse, and human genes, we show with a large number of genes that the gene duplication activity is significantly higher after the separation of amphioxus and the vertebrate lineages, which we estimate at 650 million years (Myr). The majority of human orthologs of 195 CDY groups that could be dated by the molecular clock appear to be duplicated between 300 and 680 Myr with a mean at 488 million years ago (Mya). We detected 485 duplicated chromosomal segments in the human genome containing CDY orthologs, 331 of which are found duplicated in the mouse genome and within regions syntenic between human and mouse, indicating that these were generated earlier than the human-mouse split. Model based calculations of the codon substitution rate of the human genes included in these segments agree with the molecular clock duplication time-scale prediction. Our results favor at least one large duplication event at the origin of vertebrates, followed by smaller scale duplication closer to the bird-mammalian split.

Full text

Page 1

New Evidence for Genome-Wide Duplications
at the Origin of Vertebrates Using an Amphioxus
Gene Set and Completed Animal Genomes
Georgia Panopoulou,1,4Steffen Hennig,2Detlef Groth,2Antje Krause,3
Albert J. Poustka,1Ralf Herwig,2Martin Vingron,3and Hans Lehrach1,2
1Evolution and Development Group, Department Professor H. Lehrach, Max-Planck Institut fu ¨r Molekulare Genetik, D-14195
Berlin, Germany;2Bioinformatics Group, Department Professor H. Lehrach, Max-Planck Institut fu ¨r Molekulare Genetik,
D-14195 Berlin, Germany;3Computational Molecular Biology, Department Professor M. Vingron, Max-Planck Institut fu ¨r
Molekulare Genetik, D-14195 Berlin, Germany
The 2R hypothesis predicting two genome duplications at the origin of vertebrates is highly controversial.
Studies published so far include limited sequence data from organisms close to the hypothesized genome
duplications. Through the comparison of a gene catalog from amphioxus, the closest living invertebrate relative
of vertebrates, to 3453 single-copy genes orthologous between Caenorhabditis elegans (C), Drosophila melanogaster (D),
and Saccharomyces cerevisiae (Y), and to Ciona intestinalis ESTs, mouse, and human genes, we show with a large
number of genes that the gene duplication activity is significantly higher after the separation of amphioxus and
the vertebrate lineages, which we estimate at 650 million years (Myr). The majority of human orthologs of 195
CDY groups that could be dated by the molecular clock appear to be duplicated between 300 and 680 Myr
with a mean at 488 million years ago (Mya). We detected 485 duplicated chromosomal segments in the human
genome containing CDY orthologs, 331 of which are found duplicated in the mouse genome and within regions
syntenic between human and mouse, indicating that these were generated earlier than the human–mouse split.
Model based calculations of the codon substitution rate of the human genes included in these segments agree
with the molecular clock duplication time-scale prediction. Our results favor at least one large duplication event
at the origin of vertebrates, followed by smaller scale duplication closer to the bird–mammalian split.
[Supplementary material is available online at www.genome.org. The cDNA clones used in the EST sequencing
are available from http://www.rzpd.de/. All ESTs are deposited in dbEST (accession nos. BI385298–BI388632,
BI378370–BI381823, BI381824–BI385297, and BI376198–BI378369). The consensus of the alignments of the C, D,
Y, orthologs included in the CD/CDY groups that we describe are available for similarity searches at
http://www.molgen.mpg.de/∼amphioxus.]
The eukaryotic genomes during almost 2 billion years of evo-
lution have been shaped by a large number of processes. An
understanding of these processes will be essential for unrav-
elling the multitude of biological processes encoded by the
genome. An important step in evolution seems to have taken
place at the boundary between cephalochordates with simple
body structures and the vertebrates. A plausible explanation
for the increase in phenotypic complexity at this point would
be a drastic increase in gene numbers, as it could be achieved
through the mechanism of a whole-genome duplication, a
theory introduced by Susumo Ohno (Ohno 1970). Gene du-
plications are now recognized as an important mechanism in
shaping eukaryotic genomes. However, the hypothesis pro-
posing two rounds of whole-genome duplication (2R hypoth-
esis) at the origin of vertebrates is under heavy debate (Hol-
land et al. 1994; Gibson and Spring 2000; Hughes et al. 2001).
Even after having the complete sequence of a vertebrate ge-
nome, opinions vary greatly (Friedman and Hughes 2001;
McLysaght et al. 2002). As common estimates of the evolu-
tionary time that has elapsed since the speculated whole-
genome duplications range from 400 to 500 Myr (Skrabanek
and Wolfe 1998), it is expected that if such a phenomenon
had happened, most evidence will be fragmental.
The cephalochordate amphioxus, being the closest living
invertebrate relative of vertebrates (Wada and Satoh 1994) is a
crucial organism for resolving the question of gene/genome
duplications at the origin of vertebrates (Sidow 1996; Hughes
1999). The cloning of a number of amphioxus genes in the
past (the most prominent example being that of a single Hox
cluster [Garcia-Fernandez and Holland 1994]), most of them
identified as single copy (Holland 1996), has helped form the
idea that amplification of gene numbers in vertebrates has
occurred on the vertebrate lineage after its divergence from
amphioxus (for review, see Holland 1999). However, most
studies published so far include a rather limited number of
genes from organisms evolutionarily close to the point of the
postulated two-genome duplications, and, therefore, they
consider the duplication pattern of genes for which there is
no proof showing that they were duplicated at the respective
time period. Furthermore, most of the amphioxus genes that
have been cloned and used in gene duplications studies so far
4Corresponding author.
E-MAIL panopoul@molgen.mpg.de; FAX 49-30-84131128.
Article and publication are at http://www.genome.org/cgi/doi/10.1101/
gr.874803.
Article
1056Genome Research
www.genome.org
13:1056–1066 ©2003 by Cold Spring Harbor Laboratory Press ISSN 1088-9051/03 $5.00; www.genome.org

Page 2

are mainly developmental genes that may be under different
selection constraints that could affect their duplication pat-
tern. Therefore, it is necessary to evaluate the 2R hypothesis
in light of a large and functionally broader range of genes
from amphioxus, but also taking advantage of the recently
completed human and mouse genomes. We have generated a
nonredundant catalog of genes from amphioxus (Branchio-
stoma floridae). To focus on genes that are likely to have been
duplicated during the duplication event predicted by the 2R
hypothesis, we constructed a set of groups containing or-
thologous genes that exist as single copy at least up to the
protostome–deuterostome split, namely, in the C. elegans, D.
melanogaster, and S. cerevisiae genomes. After assigning genes
from ciona, amphioxus, mouse, and human to the above or-
thologous groups, we tested the 2R hypothesis (1) by estimat-
ing the percentage of gene duplications that occurred after
the emergence of ciona and amphioxus, and (2) by examining
whether these gene duplications are due to a genome-wide
event.
RESULTS
An Amphioxus Gene Catalog and Its Complexity
A total of 14,189 5? ESTs were generated from two amphioxus
cDNA libraries prenormalized by oligonucleotide fingerprint-
ing (see Methods), which were subsequently assembled to
9173 consensus sequences (see Methods). On the basis of
sample 3? EST sequencing and the performance assessment of
the oligonucleotide fingerprinting, we expect ∼15% redun-
dancy in our gene catalog. Selection of clones for sequencing
on the basis of the oligonucleotide fingerprinting results does
not compromise the use of the EST catalog for gene-
duplication studies, as oligonucleotide fingerprinting can dis-
tinguish between members of gene
families (see Methods). A total of
22% of the amphioxus sequence
clusters show similarity to mamma-
lian and 16% to zebrafish nonre-
dundant clustered EST sets, whereas
17% and 14% show similarity to fly
and worm-predicted proteins, re-
spectively, at a BLAST E-value
<10?10. The most abundant genes
of the EST catalog are enzymes
(14.6%) and genes involved in cell
growth and maintenance (14.4%),
whereas 16.2% are intracellular and
15.7% membrane components
(Suppl. Table S1).
A Set of Invertebrate
Single-Copy Orthologous
Genes as a Means for
Counting Gene Duplications
Single gene duplications have been
proposed as an alternative to the
whole-genome duplication sce-
nario (Hughes et al. 2001). To dis-
tinguish whether the gene duplica-
tion rate was higher at the origin of
vertebrates rather than earlier in
evolution, we constructed two sets
of groups containing orthologs be-
tween organisms that have branched off prior to the genome
duplications predicted by the 2R hypothesis, for which the
sequence of their complete genome was also available. One
set of groups includes orthologs that exist as single copy in C.
elegans (C), D. melanogaster (D), and S. cerevisiae (Y), or in C.
elegans (C) and D. melanogaster (D) only (from now on referred
to as single copy CD/CDY groups); one set includes ortholo-
gous groups containing more than a single ortholog in at least
one of the compared genomes (from now on referred to as
multicopy CD/CDY groups). The imposed condition of being
single copy in all of the compared genomes ensures that the
single-copy state is ancestral rather than due to secondary
gene loss of duplicates. We use the single-copy CD/CDY
groups to identify how many of the genes that exist as single
copy, at least up to the protostome–deuterostome split, have
single amphioxus and multiple vertebrate orthologs. This is
done by attaching the amphioxus, human, and mouse or-
thologs to the platform of single-copy CD/CDY groups, and
counting the copy number of the amphioxus and vertebrate
orthologs per group.
Finally, we use the multicopy CD/CDY groups to esti-
mate the percent of single gene duplications by counting the
percent of the multicopy CD/CDY groups that contain more
orthologs from one organism (e.g., D. melanogaster) rather
than from the other two organisms involved (C. elegans and S.
cerevisiae). The CD and CDY sets of single or multicopy or-
thologous groups do not overlap.
To define the groups of orthologous genes, we used the
best hit (BeT) method (Tatusov et al. 1997), in which pairwise
comparisons (using BLAST) between all of the participating
complete genomes are performed, and the best hit(s) of each
protein in each of the other genomes in both directions is
identified as its ortholog(s). Orthologs are genes that have
Figure 1
functional classification in Gene Ontology-defined molecular function type classes. (B) Differences in
the functional classes of human orthologs according to whether they belong to single, 2–5, and 5–10
human member CD/CDY groups.
(A) Differences in the type of genes included in the CD and CDY groups on the basis of their
An Amphioxus Gene Catalog and the 2R Hypothesis
Genome Research
www.genome.org
1057

Page 3

evolved by vertical descent from a common ancestor, whereas
paralogs originate from gene duplications within a genome
(Fitch 1970). The simultaneous inter-relationship between
proteins in the compared genomes as a condition of their
assignment to the same orthologous group, a particularly
stringent criteria of assigning orthology, makes the BeT
method independent of similarity scores. This is the main
difference from other approaches that define orthologs as
those above a certain similarity threshold and/or threshold of
alignment length in reciprocal whole-genome comparisons
(Chervitz et al. 1998; Rubin et al. 2000). These two significant
methodological differences result in the identification of a
higher number of orthologous groups than reported previ-
ously, as well as to a different copy-rate distribution of or-
thologs within the groups.
We identified 1419 single-copy CDY groups and 2034
single-copy CD groups, and an additional 432 multicopy CD/
CDY groups. At least the number of single-copy CD/CDY
groups correlates to the number reported previously by Rubin
et al. (2000) (1833 yeast proteins in a set of orthologous genes
that also included 3303 fly and 3229 worm proteins). How-
ever, it is much higher than the number of orthologous
groups reported by Friedman and Hughes (2001) (they report
only 716 orthologous families between human and S. cerevi-
siae, whereas we find that 1265 CDY groups contain human
orthologs, see below).
Because it has been suggested that the preservation of
gene duplicates can vary according to the function that they
assume, we were interested in whether the CD/CDY groups
represent all functional classes of genes. We found that the
most frequent classes in both CD and CDY groups are en-
zymes, ligand binding or carriers, and transporters if classified
by Gene Ontology (Ashburner et al. 2000) molecular function
classes (Fig. 1A). This distribution is similar to the distribution
reported for the whole-human genome (Venter et al. 2001),
which indicates that the CD/CDY groups represent funda-
mental genes, and are not biased toward a specific functional
class.
Counting Gene Duplications at the Transition
From Invertebrates to Vertebrates
To estimate the extent of gene duplications at the origin of
vertebrates, we have subsequently identified the human,
mouse, ciona, and amphioxus orthologs for each of the
single-copy CD/CDY groups. We identified human orthologs
for 3044 of the single-copy CD/CDY orthologous groups (2.6
times more than reported previously [Lander et al. 2001]),
mouse orthologs for 2671 of the single-copy CD/CDY groups,
and amphioxus orthologs for 773 single-copy CD/CDY or-
thologous groups (1248 genes; Table 1A). If one or more large-
scale gene duplication event(s) have occurred at the origin of
vertebrates, one would expect that the majority of the single-
copy CD/CDY groups will contain a single amphioxus ortho-
log as opposed to multiple human and mouse orthologs. Our
results show that this is the case.
We find an average of 1.6 amphioxus gene copies per
group as compared with 4.2 for human in the 711 CD/CDY
groups that contain orthologs from both species (Table 1A).
Similar ratios are found between amphioxus and mouse.
Overall, the majority of the single-copy CD/CDY orthologous
groups contain human (62%) or mouse (56.2%) genes in mul-
tiple copies, whereas in 70.9% of the groups with an amphi-
oxus ortholog, these exist as single copy (Table 1B). The num-
ber of cases in which the opposite is true, that is, the number
of single-copy CD/CDY groups that contain multiple amphi-
oxus and single human (32 groups) or mouse (40 groups)
orthologs is statistically significantly lower, as indicated by
the paired signs test (P = 7.386e-9). A detailed inspection of
the groups shared between amphioxus/human and amphi-
Table 1A.
Mouse, or Human
Comparison of Gene Copy Number Per Single Copy CD/CDY Orthologous Group Between Amphioxus and Ciona,
Amphioxus (A)Ciona (C) Mouse (M)Human (H)
NGN/GNG N/GNGN/GNGN/G
All
(=1)
(>1)
(?1)A:(?1)C
1A:1C
1A:(>1)C
(>1)A:1C
(>1)A:(>1)C
(?1)A:(?1)M
1A:1M
1A:(>1)M
(>)A:1M
(>1)A:(>1)M
(?1)A:(?1)H
1A:1H
1A:(>1)H
(>1)A:1H
(>1)A:(>1)H
1248
548
700
945
237
128
197
383
1115
192
277
109
537
1164
180
319
106
559
773
548
225
546
237
128
75
106
676
192
277
46
161
711
180
319
46
166
1.6
1
3.1
1.7
1
1
2.6
3.6
1.6
1
1
2.4
3.3
1.6
1
1
2.3
3.4
2685
1133
1552
1094
237
330
75
452
1679
1133
546
546
237
128
75
106
1.6
1
2.8
2
1
2.5
1
4.2
7384
1173
6211
2671
1173
1498
2.8
1
4.1
10209
1156
9053
3044
1156
1888
3.4
1
4.8
2427
192
1114
46
1075
676
192
277
46
161
3.6
1
4
1
6.7
2917
180
1465
46
1226
711
180
319
46
166
4.2
1
4.6
1
7.4
The number N of genes assigned into G groups (single copy CD/CDY) and the corresponding copy ratio N/G is shown for various cases. (All)
The number of all groups containing genes from a specific organism; (>1) Groups have more than one gene from a specific organism. The
fourth (light gray), the fifth (dark gray), and the sixth row (light gray) each compare various cases for only those groups in common between
amphioxus (A) and ciona or amphioxus and mouse (M) or amphioxus and human (H).
Panopoulou et al.
1058Genome Research
www.genome.org

Page 4

oxus/mouse (Table 1A) shows clearly that the copy ratios are
correlated, that is, on an average higher gene-copy rates in
amphioxus (>1, rate 3.4) give increased gene copy rates in
human/mouse (>1, rate 7.4/6.6), which is a strong argument
for a coherent duplication mechanism like whole-genome du-
plication. Finally, it is worth stressing that the distribution of
the single-copy human orthologs in functional classes is simi-
lar to that of multicopy orthologs (Fig. 1B).
A minority of the single-copy CD/CDY groups contains
multiple amphioxus genes (225 groups, 29.1%; Table 1A).
This could be due to the estimated 15% redundancy in our
catalog (as discussed above), or the counting of splice variants
as duplicated genes, or due to misassignment of multiple 5?
ESTs representing the same transcript in the same ortholo-
gous group, or to single gene duplications that occurred either
in the common ancestor of chordates or within the amphi-
oxus lineage.
Because oligonucleotide fingerprinting recognizes splice
variants as different transcripts, we expected that a high per-
cent of the orthologous groups with multiple amphioxus or-
thologs represent alternative spliced products. However,
sample 3? EST sequencing showed that only 1 of 15 of the
orthologous groups that we tested contained splice variants.
Another misclustering effect can result from the presence of
long domains or multiple copies of short domains. However,
as 81% of the amphioxus consensus sequences included in
the single-copy CD/CDY groups have at least 100 amino acids
overlap with the rest of the genes in the same group, we ex-
pect that this has a limited effect in our analysis.
Thus, we believe that most of the multiple-copy amphi-
oxus orthologs represent true gene duplicates. To trace when
these duplications happened, we assembled a consensus EST
catalog from 88,418 publicly available ESTs of Ciona intestina-
lis (Satou et al. 2001) into 17,492 sequence clusters, which we
included in the single-copy CD/CDY groups (Tables 1A and
1B). A total of 237 (43%) of 546 CD/CDY groups that contain
orthologs from both amphioxus and ciona include 1 ortholog
from both species. A total of 75 (13%) of 546 of the CD/CDY
groups common between the two species contain 1 ciona or-
tholog, and >1 amphioxus ortholog, whereas in 40 (7.32%) of
the common CD/CDY groups, the opposite is true. Finally,
106 (19.41%) of the common groups contain more than a
single ortholog from both species. On the basis of this and our
estimation of the rate of single gene duplications via the multi
CD/CDY groups (see below), we conclude that the amphioxus
duplicates that we observe are the result of duplications that
occurred at the common ancestor of ascidians and cephalo-
chordates (this is also supported by the similarity in the du-
plication rates between amphioxus and ciona [Tables 1A and
1B]) and lineage-specific duplications.
The 432 groups with multiple copies in C, D, and Y show
little correlation in the rates of duplicates. In 89% (Y vs. C and
D) or 83% (C vs. D) of the respective common groups, extra
copies are observed only in one of the species, so that lineage-
specific single gene duplications are the most likely mecha-
nism. In total, only 5%–6% of all CD/CDY groups (single and
multicopy) contain >1 copy from either C, D, or Y, which is
very small as compared with the 44.7% of the CD/CDY groups
with a single amphioxus ortholog containing multiple hu-
man/mouse orthologs. This emphasizes that the gene dupli-
cations that occurred at the origin of vertebrates were the
result of a large-scale gene duplication event rather than
single gene duplications.
There are 2551 CD/CDY groups that contain orthologs
from both human and mouse. A total of 20% of these are
common between the two organisms’ groups and contain
more than four orthologs in both human and mouse (Table
1B). In 50% of those cases, the same number of extra copies
occurs in both mouse and human, indicating that they are the
result of duplications common to both species that predated
the human–mouse split (>100 Mya).
Evidence for a Complete Genome Duplication Versus
Alternative Mechanisms
Four criteria have been most commonly used to evaluate the
2R hypothesis.
The Number of Duplicates Per Gene Family
We find that the majority of duplicated genes in both human
(20.2% of the single-copy CD/CDY orthologous groups) and
mouse (22.2% of the single-copy CD/CDY orthologous
groups) genomes are organized in orthologous groups of two
members (Table 1B). Although this finding is in qualitative
Table 1B.
Mouse, and Human Genes They Contain
Size Distribution of the Single Copy CD/CDY Orthologous Groups According to the Number of Amphioxus, Ciona,
N/G
Amphioxus HumanAmphioxus MouseAmphioxus Ciona
G% of allG % of allG% of allG % of allG% of allG% of all
A. All group members counted
1548
2128
3 49
417
>4 31
B. Only groups common between amphioxus-human, amphioxus-mouse, and mouse-ciona are counted
149970.2% 22621.8%
2 12117.0% 13118.4%
3 456.3%9112.8%
4 162.3% 69 9.7%
>430 4.2%19427.3%
70.9%
16.6%
6.3%
2.2%
4.0%
1156
615
366
249
658
38%
20.2%
12%
8.2%
21.6%
548
128
49
17
31
70.9%
16.6%
6.3%
2.2%
4.0%
1173
592
334
167
405
43.9%
22.2%
12.5%
6.3%
15.2%
548
128
49
17
31
70.9%
16.6%
6.3%
2.2%
4.0%
1133
341
120
46
39
67.5%
20.3%
7.1%
2.7%
2.3%
469
119
44
15
29
69.3%
17.6%
6.5%
2.2%
4.3%
238
134
94
42
165
35.3%
19.8%
13.9%
6.2%
24.4%
365
103
36
15
27
66.8%
18.7%
6.6%
2.7%
4.9%
312
129
46
28
31
57.1%
23.6%
8.4%
5.1%
5.7%
The top half of the table lists the groups according to the number of genes from each organism separately; the bottom half lists the respective
ratios for the groups common between two species.
An Amphioxus Gene Catalog and the 2R Hypothesis
Genome Research
www.genome.org
1059

Page 5

agreement with recent publications (Friedman and Hughes
2001; Lander et al. 2001), the number of orthologous groups
shared between human and invertebrates that contain mul-
tiple copies in human is significantly higher in our study. As
an example, we find that 21.6% of the orthologous groups
with members from both species contain two human or-
thologs and a single S. cerevisiae ortholog, as opposed to Fried-
man and Hughes (2001), who report that 11.0% of the groups
shared between the two organisms have a 2:1 human-S. cer-
evisiae copy rate.
In addition, a high proportion of the single-copy or-
thologous groups contain a single human (38%, 1156 groups)
or mouse (43.9%, 1173 groups) ortholog. This family-size dis-
tribution is not altered if single exon genes (which comprise
7.86% of all the predicted human genes [ENSEMBL build 28]
assigned to the CD/CDY groups) that might represent pro-
cessed pseudogenes that do not result from gene duplications
are excluded (Suppl. Table S2). The presence of a large pro-
portion of human genes as single copy or in two-member
families is not in contradiction with the 2R hypothesis, as
extensive gene loss is known to happen after duplication
(Lynch and Conery 2000; Lynch 2002). In addition, we expect
that the number of single-copy genes is lower than antici-
pated at present, as silenced genes have not been included. As
an example, we found additional GENSCAN (Burge and
Karlin 1997) predicted human orthologs of single-copy CD/
CDY groups that do not exist in the ENSEMBL set (422 of 616
single-copy CD/CDY groups that have no ortholog in the
ENSEMBL set, match 3001 human GENSCAN predicted pro-
teins [April 2001, freeze]), which might represent pseudo-
genes (Suppl. Table S3). The addition of these genes can in-
crease the number of multimember orthologous groups at the
expense of the single-copy groups (Suppl. Fig. S1).
The Branching Pattern of Phylogenetic Trees of the Duplicated Genes
Although the presence of the (AB)(CD) topology has been
used as a means for evaluating the 2R theory (Burge and
Karlin 1997; Hughes 1999; Wang and Gu 2000; Hughes et al.
2001; Lander et al. 2001) or even to resolve the timing of gene
duplications (Friedman and Hughes 2001), this criterion is
highly disputed as the topology of, for example, closely timed
duplications cannot be easily resolved with the current phy-
logenetic methods (Gibson and Spring 2000). We find that
even including the amphioxus orthologs, which, because
they are evolutionarily much closer to vertebrate than the fly
or worm genes, offer better resolution in the phylogenetic tree
and avoid potential artifacts (e.g., long branch attraction),
results in very few trees (one of nine of the four-member and
three of five of the five-member groups) with the topology
expected by the 2R hypothesis. The above observation is
based on the phylogenetic trees of 62 CD/CDY groups (in-
cluding 34 amphioxus genes available from GenBank) (Suppl.
Table S4 and Fig. S4). In conclusion, the phylogenetic analysis
has resulted in findings similar to other studies, that is, al-
though a higher number of amphioxus genes was used, al-
most all of the tree topologies could agree with one round of
genome duplication, followed by additional duplications or
two rounds of genome duplications followed by gene loss.
Identification of Duplicated Segments in the Human Genome
The identification of duplicated segments in the human ge-
nome that could have been derived from a duplication event
at the origin of vertebrates has been addressed for either spe-
cific regions in the human genome such as the Hox cluster-
bearing chromosomes (Hughes et al. 2001; Larhammar et al.
2002) or the entire human genome (McLysaght et al. 2002). A
central issue in the above studies that is debatable is whether
all genes that are included in the segments are the result of the
same duplication event, and furthermore, whether this event
was the one predicted by the 2R hypothesis. The classification
of human genes to orthologous groups, which are known to
have invertebrate genes only in single copy, at least up to the
protostome–deuterostome split prior to finding whether they
are organized in duplicated segments in the human genome is
an important difference of our approach to the method of
McLysaght et. al. (2002). We identified duplicated segments
by using each chromosome as query to investigate whether
the human orthologs of a series of single-copy CD/CDY
groups on the query chromosome are also found in a specific
order and neighborhood on another chromosome (target). To
compensate for genomic rearrangements, for example, chro-
mosome inversions, translocations, tandem duplications, and
gene loss, we started from the smallest unit (pairs of CD/CDY
gene groups) that could be conserved, which we then ex-
tended to larger segments, while we also allowed for undupli-
cated genes to intervene between the duplicated CD/CDY
pairs (Fig. 2A). We found that only specific chromosomes
share CD/CDY pairs that are significant (e.g., Chromosomes 2
and 12, 1 and 6, 1 and 9, 1 and 4, etc.; Fig. 2B). Furthermore,
by comparing the frequency of the segment sizes from all
human chromosomes with those of a randomized human
chromosome set (genes positioned randomly), we found clear
significance for only those CD/CDY gene pairs that are not
separated by more than four unrelated CD/CDY groups’ genes
(Suppl. Fig. S3). Therefore, we excluded from further analysis
all pairs with more than four intervening genes that belong to
other than the query CD/CDY pairs. As each CD/CDY group
gene is separated, on average, by 1.3 genes that are not related
to a CD/CDY group, this in extent means that the maximum
number of genes allowed between CD/CDY pairs is around 10
genes (4 CD/CDY group genes plus 1.3 ? 5 = 6.5 non-CD/
CDY genes). Thus, as opposed to McLysaght et al. (2002), who
allow for 30 unduplicated genes between 2 duplicated genes
in each paralogon, we found that much less intervening un-
related genes can be allowed if one wants to avoid inclusion of
segments that contain duplicated genes found as neighbor by
chance. We finally extended the identified pairs of genes by
joining neighbor pairs that had one CD/CDY group in com-
mon. Overall, we detected 1872 CD/CDY group gene pairs
(961 of which had zero intervening unduplicated CD/CDY
group genes) containing 2511 genes, when 0–4 unrelated CD/
CDY group genes are allowed. The above 1872 CD/CDY hu-
man gene pairs (or pair combinations) are included in 656
CD/CDY groups. A total of 192 of these CD/CDY groups with
a human ortholog contained within a detected duplicated
segment also include an amphioxus ortholog. Neighbor mini-
mal size units (pairs) were subsequently joined in 485 larger
segments (extended segments) containing 1611 genes (Table
2). As an example, 21 segments are shared between chromo-
somes 2 and 12, two of the most frequently quoted chromo-
somes (Hox-containing chromosomes) in gene duplication
studies (Fig. 4). The largest duplicated segment containing 5
genes is located on 1 (46.119 Mb), with its duplicon on chro-
mosome 12 containing 6 genes (5.519Mb). However, the ma-
jority of the extended segments contain 2–3 CD/CDY group
genes (Table 2).
A first rough estimate of the age of these duplicated seg-
ments can be performed by using the known synteny rela-
Panopoulou et al.
1060 Genome Research
www.genome.org