Genome desertification in eutherians: can gene deserts explain the uneven distribution of genes in placental mammalian genomes?

Abstract

The evolution of genome size as well as structure and organization of genomes belongs among the key questions of genome biology. Here we show, based on a comparative analysis of 30 genomes, that there is generally a tight correlation between the number of genes per chromosome and the length of the respective chromosome in eukaryotic genomes. The surprising exceptions to this pattern are placental mammalian genomes. We identify the number and, more importantly, the uneven distribution of gene deserts among chromosomes, i.e., long (>500 kb) stretches of DNA that do not encode for genes, as the main contributing factor for the observed anomaly of eutherian genomes. Gene-rich placental mammalian chromosomes have smaller proportions of gene deserts and vice versa. We show that the uneven distribution of gene deserts is a derived character state of eutherians. The functional and evolutionary significance of this particular feature of eutherian genomes remains to be explained.

Full text

Page 1

Genome Desertification in Eutherians: Can Gene Deserts
Explain the Uneven Distribution of Genes in Placental
Mammalian Genomes?
Walter Salzburger Æ Æ Dirk Steinke Æ Æ Ingo Braasch Æ Æ
Axel Meyer
Received: 5 May 2009/Accepted: 15 May 2009/Published online: 1 July 2009
? Springer Science+Business Media, LLC 2009
Abstract
ture and organization of genomes belongs among the key
questions of genome biology. Here we show, based on a
comparativeanalysisof30genomes,thatthereisgenerallya
tight correlation between the number of genes per chro-
mosome and the length of the respective chromosome in
eukaryotic genomes. The surprising exceptions to this pat-
tern are placental mammalian genomes. We identify the
The evolution of genome size as well as struc-
number and, more importantly, the uneven distribution of
gene deserts among chromosomes, i.e., long ([500 kb)
stretches of DNA that do not encode for genes, as the main
contributing factor for the observed anomaly of eutherian
genomes. Gene-rich placental mammalian chromosomes
have smaller proportions of gene deserts and vice versa. We
showthattheunevendistributionofgenedesertsisaderived
character state of eutherians. The functional and evolu-
tionary significance of this particular feature of eutherian
genomes remains to be explained.
Keywords
Genome size ? Mammalia
Genome evolution ? Gene deserts ?
Abbreviations
kb
LINE
SINE
TE
Kilobase
Long interspersed nuclear element
Short interspersed nuclear element
Transposable element
Introduction
One of the major challenges of genome biology is the
understanding of how genomes are organized and how
genes are distributed in genomes of different sizes. This
question was brought to the fore and attracted interest after
it was discovered that genomes of vastly different sizes
often contain relatively similar numbers of genes (Bork and
Copley 2001; Gregory 2005; Lynch 2007). Recent com-
parative genomic studies have uncovered some general
trends in the evolution of genome size. For example, it
appears that larger genomes contain proportionally fewer
genes compared with smaller ones (see, e.g., Gregory
2005); however, they are characterized by higher numbers
W. Salzburger, D. Steinke and I. Braasch contributed equally to this
work.
Electronic supplementary material
article (doi:10.1007/s00239-009-9251-4) contains supplementary
material, which is available to authorized users.
The online version of this
W. Salzburger ? D. Steinke ? I. Braasch ? A. Meyer (&)
Lehrstuhl fu ¨r Zoologie und Evolutionsbiologie,
Department of Biology, University Konstanz,
Constance 78457, Germany
e-mail: axel.meyer@uni-konstanz.de
Present Address:
W. Salzburger
Zoological Institute, University of Basel,
Basel 4051, Switzerland
e-mail: walter.salzburger@unibas.ch
Present Address:
D. Steinke
Biodiversity Institute of Ontario, University of Guelph,
Guelph, Ontario N1G 2W1, Canada
e-mail: dsteinke@uoguelph.ca
Present Address:
I. Braasch
Physiological Chemistry I, University of Wu ¨rzburg,
97074 Wu ¨rzburg, Germany
e-mail: ingo.braasch@biozentrum.uni-wuerzburg.de
123
J Mol Evol (2009) 69:207–216
DOI 10.1007/s00239-009-9251-4

Page 2

of transposable elements (TEs; Kidwell 2002; Lynch and
Conery 2003). It has further been shown that the total
number of genes in a genome is strongly correlated with
the length of the protein-coding sequence in both pro-
karyotic and eukaryotic genomes, suggesting that gene
length is highly conserved within each of these two
anciently diverged lineages (Xu et al. 2006, which includes
most of the taxa reported herein). However, whereas gene
number and genome size are correlated in prokaryotes
(Gregory and DeSalle 2005), such a trend does not exist in
eukaryotic genomes, in which only a small fraction of the
nuclear DNA is protein-coding. Finally, the whole genome
sequencing of mammals during the last years (see, e.g.,
Lander et al. 2001; Lindblad-Toh et al. 2005; Venter et al.
2001) showed that genes are not evenly distributed in a
given genome and that substantial fractions of mammalian
genomes— B25% in the case of Homo sapiens—are made
up of so-called gene deserts, i.e., long regions[500 kb in
length that are devoid of any genes (Venter et al. 2001).
Here we show that in all eukaryotic genomes analyzed,
except in placental mammalian ones, the number of genes
per chromosome is strongly correlated with chromosome
length, irrespective of genome size, chromosome number,
and taxonomy. We then tested whether particular genomic
features, such as repetitive elements or gene deserts, may
account for this difference. We found that the distinctive-
ness of placental mammal genomes can be best explained
by the uneven distribution of gene deserts.
Materials and Methods
Data Mining
WeobtainedchromosomelengthdatafromtheUniversityof
California Santa Cruz (UCSC) genome browser (Karolchik
etal.2003)andtheEuropeanMolecularBiologyLaboratory
(EMBL) EuropeanBioinformaticsInstitute(EBI)Web site(
http://www.ebi.ac.uk). Coding genes and their number by
chromosome were obtained from the EBI website for the
following species chosen to be representative for the major
lineage of organismal diversity (in alphabetical order):
Arabidopsis thaliana (thale cress), Aspergillus fumigatus,
Bos taurus (domestic cow), Caenorhabditis briggsae
(roundworm), Candida glabrata (candida yeast), H. sapiens
(human), Leishmania major, Mus musculus (house mouse),
Oryza sativa (domestic rice), Ostreococcus lucimarinus,
Plasmodium falciparum,
(baker’syeast),andVitisvinifera(commongrapewine).We
also used the UCSC genome browser for data from Anoph-
eles gambiae (mosquito), C. elegans (roundworm), Canis
lupus familiaris (dog), Ciona intestinalis (vase tunicate),
Saccharomycescerevisiae
Daniorerio(zebrafish),Drosophilamelanogaster(fruitfly),
Equus caballus (horse), Gallus gallus (chicken), Gasteros-
teus aculeatus (three-spine stickleback), Macaca mulatta
(rhesus monkey), Monodelphis domestica (gray short-tailed
opossum), Ornithorhynchus anatinus (platypus), Oryzias
latipes (Japanese killifish), Pan troglodytes (chimpanzee),
Rattus norvegicus (brown rat) and Tetraodon nigroviridis
(green spotted puffer fish). In addition we used National
Center for Biotechnology Information GenBank and the
Maize Genetics and Genomics Database for estimations for
Zeamays(maize)(http://www.maizegdb.org).Alldatawere
downloaded in March 2008.
Data for the number of TEs, long interspersed nuclear
elements (LINEs), short interspersed nuclear elements
(SINEs), and simple repeats were obtained from the UCSC
genome browser (Karolchik et al. 2003) for A. gambiae, B.
taurus, C. briggsae, C. elegans, C. lupus familiaris, C.
intestinalis, D. rerio, D. melanogaster, E. caballus, G.
gallus, G. aculeatus, H. sapiens, M. mulatta, M. domestica,
M. musculus, O. anatinus, O. latipes, P. troglodytes, R.
norvegicus, and T. nigroviridis, and from the Arabidopsis
information resource (i.e., TAIR) for A. thaliana. All data
were downloaded in March 2008.
We used the information provided by the UCSC genome
browser to identify intergenic regions in A. gambiae, B.
taurus, C. briggsae, C. elegans, C. lupus familiaris, C.
intestinalis, D. rerio, D. melanogaster, E. caballus, G.
gallus, G. aculeatus, H. sapiens, M. mulatta, M. domestica,
M. musculus, O. anatinus, O. latipes, P. troglodytes, R.
norvegicus, S. cerevisiae, T. nigroviridis, and V. vinifera.
Data for P. falciparum were obtained from the Plasmodium
genome database (Kissinger et al. 2002). No information
on the size of intergenic regions was available for A. tha-
liana, A. fumigatus, L. major, O. sativa, and Z. mays.
However, it has already been shown that the larger genome
of some plant species (e.g., Z. mays) is due to the expansion
in number of transposable elements (Kidwell 2002;
Messing et al. 2004). We follow the definition provided by
Venter et al. (2001), who classified all intergenic regions
[500 kb as gene deserts (see Table 1 for a list of taxa,
their genome sizes, and the number and percentage fraction
of gene deserts). We note that other investigators have
applied a modified classification in mammals, defining only
the 3% longest intergenic intervals as gene deserts (Ov-
charenko et al. 2005). However, this strategy was not
suitable for our approach comparing a variety of organisms
because it would have led to the classification of very short
intergenic regions as gene deserts in nonvertebrate gen-
omes (e.g., 12 to 70 kb in C. elegans or 2.5 to 80 kb in S.
cerevisiae). In addition, this approach would require an
a priori acceptance of the existence of gene deserts in any
given genome.
208J Mol Evol (2009) 69:207–216
123

Page 3

Analysis of Genome Data
To test the hypothesis that gene number and chromosome
length are correlated, we first plotted the number of genes
per chromosome (NG) versus chromosome length (LC) for
the genomes mentioned previously. Some limitations exist
in the inference of genome and chromosome sizes based
only on sequence data (Gregory 2005), which is in part
caused by the fact that genome sequences are rarely com-
plete. However, these slight differences in completeness
among the sequenced genomes should not affect our
analyses. We used only genomes for which a continuous
genome assembly, in the form of individual chromosomes,
was available; therefore, a satisfactory coverage of these
genomes can be assured. In addition, no systematic a priori
bias in genome completeness can be assumed. We used the
square of the correlation coefficient (R2) to describe the
goodness-of-fit of the data to the hypothesized correlation
between NGand LC. In addition, we performed pairwise
comparison using Tukey-Kramer method for unplanned
comparisons among a set of regression coefficients to
identify those pairs of genomes that show significantly
different correlations of NG/LC.
We then plotted the total number of TEs, LINEs, SINEs,
simple repeats, and gene deserts against genome size for
those organisms for which these data were available. The
same procedure was followed with the numbers of TEs,
LINEs, SINEs, simple repeats, and gene deserts per chro-
mosome. To further evaluate the contribution of gene
deserts to the distribution of genes on chromosomes in
mammalian genomes, we plotted the relative proportion of
genes per chromosome (NG/TG) plus the relative propor-
tion of gene deserts (ND/TD) versus chromosome length
(LC), where NGis the number of genes per chromosome;
TGis the total number of genes in a genome; NDis the
number of gene deserts per chromosome; and TDis the
total number of gene deserts in a genome. For mammals,
we also plotted the length-corrected sum of the relative
Table 1 Organisms used in this
study and information about
their genomesa
Deserts (n)—number of gene
deserts ([500 kB) in a given
genome; deserts (%)—size
fraction of gene deserts
([500 kB) in a given genome;
18S acc. no
aGenBank accession numbers
of 18S sequences used for
regression equation mapping are
also given. Data were obtained
from GenBank. Note that no
information on the size of
intergenic regions was available
for A. thaliana, A. fumigatus, L.
ajor, O. sativa, and Z. mays.
Also note that assembly size
does not necessarily equal
actual genome size
TaxonAssembly size (Mb)Deserts (n) Deserts (%)18S GenBank accession number
P. troglodytes
3175.694936.80 AC183378
H. sapiens
3047.091538.33NR_003286
M. mulatta
2863.781030.44CN805008
M. musculus
2654.989534.26NR_003278
R. norvegicus
2718.9 74323.19X01117
C. lupus familaris
2445.155220.14 DQ287955
B. taurus
2422.91494.26DQ222453
E. caballus
2367.1 573 23.13AJ311673
M. domestica
3431.4109827.77 AJ311676
O. anatinus
1843.0 1618.37AJ311679
G. gallus
1031.9 200 16.13AF173612
D. rerio
1277.1 1297.10 XR_045186
O. latipes
724.2 29 6.06 AB105163
G. aculeatus
400.90 0.00DW607648
T. nigroviridis
217.45 1.17AJ270032
C. intestinalis
938.10 0.00 AB013017
A. gambiae
228.20 0.00AM157179
D. melanogaster
120.400.00 EU188739
C. elegans
100.30 0.00AY284652
C. briggsae
91.20 0.00U13929
A. fumigatus
29.400.00 NT_166520
C. glabrata
12.300.00AF114470
S. cervisiae
12.100.00J01353
Z. mays
1979.4 00.00NC_008332
A. thaliana
119.200.00 X16077
O. sativa
370.800.00AY120865
V. vinifera
303.100.00AF321270
O. lucimarinus
13.200.00DQ007077
L. major
32.800.00NC_007268
P. falciparum
22.90 0.00NC_004325
J Mol Evol (2009) 69:207–216 209
123

Page 4

proportion of genes per chromosome plus the relative
proportion of gene deserts ((NG/TG? ND/TD)/LC) for each
chromosome. In addition, to test whether the number of
LINEs, SINEs, LTRs, DNA transposons, simple repeats, or
gene deserts compensates best for varying gene densities in
mammalian chromosomes, we performed a partial regres-
sion analysis based on NGand LC.
Regression Equation Mapping
We mapped the number of gene deserts as well as their
relative proportion in the respective genome onto a phy-
logeny based on 18S rRNA sequences (see Table 1 for
GenBank accession numbers) that was in concordance with
recent phylogenomic studies (see, e.g., Delsuc et al. 2006;
Dunn et al. 2008; Lartillot et al. 2007). We performed a
maximum likelihood analysis and 100 maximum likeli-
hood bootstrap replicates with PHyML (Guindon and
Gascuel 2003) using the TrN ? C model of sequence
evolution according to ModelGenerator (Keane et al.
2006). To determine whether there is a correlation between
phylogenetic position and number and relative proportion
of gene deserts, we used an independent contrast analysis
(Garland and Ives 2000) as implemented in the phenotypic
diversity analysis program in the Mesquite package
(Maddison and Maddison 2004).
Results
The plot of the number of genes per chromosome against
chromosome length showed very strong correlations for
nonmammalian genomes as well as for platypus (mono-
tremes) and opossum (marsupials) (Fig. 1), whereas pla-
cental mammalian (eutherian) genomes deviate from this
eukaryote-wide trend (Fig. 2). R2values for the plot of NG/
LCtypically are approximately C0.9 in noneutherian gen-
omes, whereas they are approximately B0.6 in mammals.
There are only few exceptions to this pattern: The only two
nonmammalian species with R2\0.6 are medaka (O.
latipes; R = 0.48) and zebrafish (D. rerio; R = 0.37).
These relatively low R2values appear to be due to the
relative equal size of chromosomes in O. latipes and by
two outlier chromosomes in D. rerio (if those two outlier
chromosomes are removed in D. rerio, R2is \0.76). The
two nematodes (C. briggsae and C. elegans) also have
relatively small R2values, which might be due to their
small number of chromosomes and the small range of sizes
of such. The only placental taxon showing R2[0.6 is the
rat (R. norvegicus; R2= 0.70).
The trend of linear correlation of NG/LCin nonplacental
mammals but not in placental mammals was further sup-
ported bythepairwise comparison ofregression
coefficients by means of Tukey-Kramer method for
unplanned comparisons. This test showed significant dif-
ferences in most comparisons between eutherian and
noneutherian genomes. The exceptions concerned all
pairwise comparisons with both Caenorhabditis species
and R. norvegicus as well as some comparisons involving
D. rerio and O. latipes. None of the pairwise comparisons
between two noneutherian (with the exception of D. rerio
and O. latipes as explained previously) or two eutherian
genomes showed significant differences in their regression
coefficients, thus pointing to a deviation from a linear
relation between NGand LConly in eutherian genomes.
Repetitive elements could be one factor to explain the
nonlinearity of NG/LCin mammals. However, we found that
the number of TEs, LINEs, and SINEs in a genome is cor-
relatedwithgenomesizebutnotwiththenumberofgenesin
a particular genome. The R2values for the plots of TEs,
LINEs, SINEs, and long terminal repeats (LTRs) against
genome size were 0.93, 0.84, 0.87, and 0.82, respectively.
When plotted against the number of genes, the R2values
were all\0.04. The numbers of TEs, LINEs, and SINEs per
chromosome also correlate with chromosome length. For
example, regarding the human genome—the most complete
and best-assembled genome of all—the corresponding R2
valueswere0.95forTEs,0.96forLINEs,and0.80forSINEs
(Supplementary Fig. 1). In addition, the number of DNA
transposons (R2= 0.94), LTRs (R2= 0.94), and simple
repeats (R2= 0.97) per chromosome was correlated with
chromosome size. Thus, because each class of repetitive
elements correlates with chromosome length, none of these
genomic features seems to account for the uneven gene
density on eutherian chromosomes.
The situation appears different when gene deserts
(intergenic regions [500 kb in size) are considered. That
gene deserts counterbalance the number of genes on
eutherian chromosomes is best illustrated by plotting the
sum of the proportion of genes per chromosome plus the
proportion of gene deserts per chromosome against chro-
mosome size ((NG/TG? ND/TD)/LC) (Fig. 3). This fac-
toring of (ND/TD) into the equation (NG/TG/LC) leads to
strong correlations in placental genomes with R2values
between 0.71 (B. taurus) and 0.98 (R. norvegicus). That
gene deserts counterbalance gene densities is further sup-
ported by the observation that the length-corrected sum of
the relative proportion of genes per chromosome plus the
relative proportion of gene deserts ((NG/TG? ND/TD)/LC)
appears relatively constant for each chromosome, except
for the Y chromosome (data not shown). Most importantly,
the partial regression analysis showed that in the genomes
of placental mammals, gene deserts, together with SINEs,
have the highest partial regression coefficients with highly
significant p values (\0.01) (Table 2). This suggests that of
all the genomic features analyzed, gene deserts are the ones
210 J Mol Evol (2009) 69:207–216
123

Page 5

1020 304050 60 70800
0
500
1000
1500
2000
2500
3000
3500
100 20
30
40
50 6070
0
1000
2000
3000
4000
5000
6000
01
23456
0
400
800
1200
1600
050
100
150200 250300
350
0
2000
4000
6000
8000
0
100200
chromosome length (L ) [Mb]
300400 500600
700
800
0
1000
2000
3000
4000
5000
0 0.5
11.522.53
3.5
0
200
400
600
800
number of genes per chromosome (NG)
Protists
Leishmania major
2
(R = 0.98)
Plasmodium falciparum
2
(R = 0.99)
Plants
Zea mays
2
(R = 0.83)
Arabidobsis thaliana
2
(R = 0.99)
Oryza sativa
2
(R = 0.89)
Vitis vinifera
2
(R = 0.99)
Ostreococcus lucimarinus
2
(R = 0.99)
Fungi
Saccharomyces cerevisiae
2
(R = 0.99)
Aspergillus fumigatus
2
(R = 0.99)
Candida glabrata
2
(R = 0.99)
Ecdysozoa
Caenorhabditis elegans
2
(R = 0.65)
Anopheles gambiae
2
(R = 0.96)
Drosophila melanogaster
2
(R = 0.95)
Caenorhabditis briggsae
2
(R = 0.75)
Actinopterygii
Monotremata, Marsupialia, Sauropsida
chromosome length (L ) [Mb]
C
Monodelphis domestica
2
(R = 0.95)
Gallus gallus
2
(R = 0.97)
Ornithorhynchus anatinus
2
(R = 0.94)
Oryzias latipes
2
(R = 0.48)
Gasterosteus aculeatus
2
(R = 0.94)
Danio rerio
2
(R = 0.37)
Tetraodon nigroviridis
2
(R = 0.92)
Urochordata
500
1000
0
0
2468
Ciona intestinalis
2
(R = 0.97)
number of genes per chromosome (NG)
number of genes per chromosome (NG)
number of genes per chromosome (NG)
number of genes per chromosome (NG)
number of genes per chromosome (NG)
C
chromosome length (L ) [Mb]
C
chromosome length (L ) [Mb]
C
chromosome length (L ) [Mb]
C
chromosome length (L ) [Mb]
C
Fig. 1 The relationship between the number of genes per chromo-
some (NG) over chromosome length (LC) shows a strong correlation
in nonmammals, irrespective of genome size, chromosome number
and taxonomy. In noneutherian genomes, the slope of the trend-line
can be interpreted as measurement for genome-compactness. The
small R2-value in zebrafish (Danio rerio; 0.37) can be explained by
two outlier chromosomes, that of medaka (Oryzias latipes; 0.48) by
the relatively equal size of its chromosomes and variance in the
number of genes. The somewhat smaller R2-value observed in
Caenorhabditis elegans (R2= 0.65) is due to the relatively even
length of its chromosomes (Nelson et al. 2004), which makes it
difficult to test for a linear relationship between NGand LC
J Mol Evol (2009) 69:207–216211
123

Page 6

that best account for and contribute to the large variation in
gene densities among mammalian chromosomes. In some
genomes, LINEs and simple repeats also showed p values
\0.01.
The mapping of the number of gene deserts as well as
their relative proportion in the respective genome onto the
phylogeny showed a substantial expansion of the number
of gene deserts in the lineage leading to the placental
mammals. Independent contrast analysis showed that the
observed trend is statistically significant (p[0.01).
Discussion
We first tested—in 30 plant, fungal, and animal genomes
for which this information was available—whether or not a
linear correlation exists between NGand LCon which they
are located. When plotting the number of genes per chro-
mosome versus chromosome length, we observed a strong
linear correlation in eukaryotic genomes, with the notable
exception of placental mammals (Fig. 1). R2values were
typically[0.9, suggesting relatively constant ratios of gene
number per chromosome length for all chromosomes in a
given genome. However, the genomes of placental mam-
mals did not show constant chromosomal gene densities
(Fig. 2), and R2values were much lower and only ranged
from 0.28 to 0.70.
The difference in NG/LC between nonplacental and
placental genomes is further substantiated by a pairwise
comparison of regression coefficients by means of Tukey-
Kramer method for unplanned comparisons, which showed
significant differences in most pairwise comparisons
between mammalian and nonmammalian genomes. The
exceptions concerned all pairwise comparisons with both
Caenorhabditis species and R. norvegicus as well as some
comparisons involving D. rerio and O. latipes. This could
be explained by the similar size of the chromosomes in O.
latipes as well as in both Caenorhabditis species (which
also seems to be responsible for the comparably low R2
values in the plot of NG/LCof 0.65 and 0.75); the relatively
small and evenly distributed gene deserts in R. norvegicus
(leading to the highest R2value of 0.70 among mammals);
and two outlier chromosomes in D. rerio. Importantly,
none of the pairwise comparisons between two nonplac-
ental or placental genomes showed significant differences
in their regression coefficients, which strongly points to a
0
1000
2000
3000
0100200300
0100 200300
2
(R = 0.39)
0
1000
2000
3000
0 100200300
0 100 200300
0
1000
2000
3000
0100200300
0
1000
2000
3000
0
1000
2000
3000
Homo sapiens
Pan troglodytesMacaca mulatta
Mus musculusRattus norvegicus
0 100200300
0
1000
2000
3000
Bos taurus
0100200 300
0
1000
2000
3000
Equus caballus
0 100200300
0
1000
2000
3000
Canis lupus
2
(R = 0.39)
2
(R = 0.33)
2
(R = 0.53)
2
(R = 0.70)
2
(R = 0.28)
2
(R = 0.57)
2
(R = 0.59)
0
1000
2000
3000
0 100 200300
number of genes per chromosome (N )
G
chromosome length (L ) [Mb]
C
Fig. 2 The correlation between NGand LCis weak in genomes of
placental mammals. In general, larger chromosomes also tend to have
more genes in mammals; however, many chromosomes significantly
deviate from a constant NG/LCratio, rendering the genome-wide trend
much weaker in placental mammalian genomes than in all other
genomes. The highest R2value in a mammal was found for rat
(R2= 0.70) whose genome contains the smallest relative fraction of
gene deserts
212 J Mol Evol (2009) 69:207–216
123

Page 7

deviation from a linear relation between NGand LConly in
eutherian genomes.
We hypothesized that this characteristic in the genomes
of placental mammals might be due to particular genomic
features of this group. We therefore examined whether the
distribution of repeats, TEs, subclasses thereof (LINEs,
SINEs), or gene deserts might be responsible for the
unexpected variation in gene densities on different mam-
malian chromosomes. To this end, we plotted the number
of TEs, LINEs, SINEs, simple repeats, and gene deserts
against genome size and chromosome length. We found
that although the number of TEs, LINEs, and SINEs is
strongly correlated with genome size itself (as already
shown by, e.g., Kidwell 2002; Lynch and Conery 2003)
and also with chromosome length, none of these classes of
repeat elements contributes particularly strongly to the
observed pattern (Supplementary Fig. 1). Instead, it
appears that the uneven distribution of gene deserts on
mammalian chromosomes accounts for the deviations from
otherwise constant ratios of NG/LC.
The plot of the sum of the relative proportion of genes
per chromosome plus the relative proportion of gene
deserts per chromosome versus chromosome length
showed a strong correlation in placental mammals, with R2
2
(R = 0.88)
Homo sapiens
Pan troglodytesMacaca mulatta
Mus musculusRattus norvegicus
Bos taurus
Equus caballusCanis lupus
0100200300
0
0.05
0.1
0
0.05
0.1
0100200300
0100200300
0
0.05
0.1
0 100200 300
0
0.05
0.1
0100200300
0
0.05
0.1
0
0.05
0.1
0100 200300
0100200300
0
0.05
0.1
0100 200300
0
0.05
0.1
2
(R = 0.79)
2
(R = 0.95)
2
(R = 0.92)
2
(R = 0.98)
2
(R = 0.71)
2
(R = 0.72)
2
(R = 0.90)
0
0.05
0.1
0100200300
relative proportion of genes and gene deserts
chromosome length (L ) [Mb]
C
Fig. 3 Gene deserts counterbalance the number of genes on mammalian chromosomes. The sum of the proportion of genes per chromosome plus
the proportion of gene deserts per chromosome is plotted against chromosome length
Table 2 Results from the partial regression analysisa
Feature
H. sapiens P. troglodytes M. mulatta M. musculusR. norvegicusC. familiarisB. taurusE. caballus
LINEs0.3300.2600.1560.0810.149-0.441*0.537*0.132
SINEs
0.618* 0.901*
0.322
0.743*0.816*
0.832*0.666*0.3562*
LTRs-0.0218 0.0340.001 0.452 0.121-0.095-0.620-0.2172
DNA transp.-0.233-0.3030.001 0.1320.1640.1400.3950.305
Simple repeats-0.132-0.1560.127-0.3740.041-0.339*-0.542*0.186
Gene deserts0.533*0.883*
0.532*
0.353 0.603*
0.837* 0.941*0.513*
aThe partial regression coefficients for the respective contribution to NG/LCis given for LINEs, SINEs, LTRs, DNA transposons, simple repeats,
and gene deserts. The highest coefficient for each genome is shown in bold, and significant values (p[0.01) are marked with an asterisk
J Mol Evol (2009) 69:207–216213
123

Page 8

values ranging from 0.71 (cow) to 0.98 (rat) (Fig. 3). This
suggests that the distribution of genes and gene deserts
counterbalance one another and predicts that, in eutherians,
chromosomes with fewer genes have proportionally more
gene deserts and vice versa. Indeed, such an observation
has already been reported from the human genome, in
which the proportion of gene deserts in the gene-rich
chromosomes 17, 19, and 22 is less than half compared
with the gene-poor chromosomes 4, 13, and 18 (Venter
et al. 2001). In addition, partial regression analysis showed
that the number of gene deserts, together with SINEs,
contribute most to the lack of fit between the number of
genes and the length of the respective chromosomes in
placental mammals (Table 2). Note that in some genomes,
LINEs and simple repeats also showed p values \0.01.
This is not surprising, given that SINEs, LINEs, and simple
repeats are not completely independent from gene deserts,
which have been shown to be enriched with such repetitive
elements (Ovcharenko et al. 2005). However, because
these repetitive elements are in general evenly distributed
across genomes (see Supplementary Fig. 1 for human),
they are unlikely to account for the uneven distribution of
gene numbers on eutherian chromosomes.
A randomization test for phylogenetic signal showed that
both the number of gene deserts and their relative propor-
tion in a genome are significantly associated with the
organisms’ phylogenetic position (p\0.01). This suggests
that the genomic organization of placental mammals,
characterized by a much higher number and an uneven
distribution of gene deserts, is a derived state among the
studied genomes. The number of gene deserts in a genome
is most likely dependent on genome size, and it will be
interesting to see how many gene deserts can be identified in
very large genomes, such as lungfish or salamander, and
whether per-chromosome gene densities also vary in these
genomes. As we show here, the distribution of gene deserts
does not appear to be dependent on genome size. This is
best illustrated by the genome sizes of platypus (O. anati-
nus; 3.0 Gb) and gray short-tailed opossum (M. domestica;
3.4 Gb), which show strong correlations between NG/LC
(R2= 0.94 and R2= 0.95, respectively) and which lie
within the range of the sizes of the placental mammalian
genomes analyzed here ranging from 3.1 Gb (horse and
dog) to 3.6 Gb (chimpanzee). Hence, the uneven distribu-
tion of gene deserts is the most likely explanation for the
deviation from constant chromosome gene density in pla-
cental mammalian genomes. We are aware that, thus far,
only a limited number of genomes are available for such
kinds of analyses. It remains to be elucidated whether or not
the uneven distribution of genes and gene deserts is also
found in other large noneutherian genomes, such as sala-
manders of lungfishes. The inclusion of larger genomes
from other lineages seems crucial to test our hypothesis.
The uneven distribution of gene deserts is not the only
peculiarity of eutherian genomes. The eutherian karyotype
seems to be extensively rearranged compared with the
ancestral vertebrate karyotype (Ferguson-Smith and Tri-
fonov 2007). This genomic rearrangement occurred after
the divergence from marsupials because the genome of the
opossum is more syntenic to the chicken genome than it is
to the human one (Ferguson-Smith and Trifonov 2007;
Mikkelsen et al. 2007). Our results, which demonstrate the
similarity of overall genome organization found in chicken,
platypus, and opossum (again to the exclusion of eutherian
mammals) are in agreement with this previous finding.
Although such chromosome rearrangements might explain
the overall similarity in the organization of placental
mammalian genomes, they cannot explain the origin of the
uneven distribution of gene deserts therein. Fusion or fis-
sion of ancestral chromosomes with an even distribution of
gene deserts along these chromosomes would necessarily
lead to an even distribution of gene deserts in the newly
rearranged chromosomes. Within-genome differences in
chromosome length do not seem to contribute to the dis-
tribution of gene deserts, either: Micro-chromosomes (as
observed in chicken or platypus) or giant chromosomes
(such as chromosome 1 in the gray short-tailed opossum)
show strong correlations in chromosome gene densities.
More than 20 years ago, Ohno (1985) postulated the
desertification of the euchromatic region of the higher
vertebrates’ genomes owing to continuous gene duplication
events followed by degeneration of newly emerged gene
copies in their evolutionary history. Only with the first
release of the complete sequence of the human genome in
2001 was Ohno’s prediction of the existence of such
deserts confirmed (Lander et al. 2001; Venter et al. 2001).
Another mechanism to generate imbalance of gene deserts
between chromosomes might be large intra-chromosomal
duplications, from which some mammalian-specific gene
deserts appear to have evolved (Itoh et al. 2005). However,
the functional or evolutionary significance of gene deserts
is not yet fully understood. At first glance, gene deserts
seem to be devoid of biologic functions because of their
lack of protein-coding DNA. Despite this, some gene
deserts have been shown to contain important regulatory
and sometimes ultraconserved regions for neighboring
genes that function over large distances (Bejerano et al.
2004; Nobrega et al. 2003; Sandelin et al. 2004). In addi-
tion, the observation of stable gene deserts with homolo-
gous flanking genes (often transcription factors), which are
maintained for long evolutionary durations (Ovcharenko
et al. 2005), as well as the existence of numerous conserved
nongenic sequences in mammalian genomes (Dermitzakis
et al. 2005; Siepel et al. 2005) suggest that gene deserts are
not just genomic junkyards but instead might be of func-
tional significance. However, some gene deserts can be
214J Mol Evol (2009) 69:207–216
123

Page 9

deleted without noticeable phenotypic effects (Nobrega
et al. 2004). Thus, it remains unclear whether the uneven
accumulation of gene deserts in eutherian chromosomes,
which appears to be the strongest causal agent for the
observed relative lack of a NG/LCrelation in the genomes
of placental mammals, is simply a byproduct of their
genome evolution and possibly the long-term decrease in
population-size (Lynch and Conery 2003), as would be
suggested by the enrichment of gene deserts along the
evolutionary lineage leading to mammals. Upcoming
detailed reconstructions of ancestral vertebrate genomes
will help clarify these points.
Alternatively, this particular architectural feature of
eutherian genomes might be linked to some of their mor-
phologic, physiologic, neurologic, and cognitive evolu-
tionary innovations, possibly by regulating genes with
essential functions in development (e.g., see de la Calle-
Mustienes et al. 2005; Taylor 2005). It has recently been
hypothesized that regulatory elements in gene deserts
function in the regulation of core vertebrate genes (Bejer-
ano et al. 2004; de la Calle-Mustienes et al. 2005; Lind-
blad-Toh et al.2005; Taylor
approximately 20% of conserved noncoding elements are
eutherian-specific (Mikkelsen et al. 2007). Thus, the
uneven distribution of gene deserts could itself be caused
by an underlying pattern of uneven distribution of some
core genes in placental genomes. This hypothesis should be
tested in the future.
2005).Furthermore,
Acknowledgments
Landesstiftung Baden-Wu ¨rttemberg GmbH (W. S.), the Center for
Junior Research Fellows of the University of Konstanz (W. S.), the
National Sciences and Engineering Research Council of Canada (D.
S.), and the German Science Foundation (A. M.). We also thank two
anonymous reviewers and I. Nanda for valuable comments.
This study was supported by grants from the
References
Bejerano G, Pheasant M, Makunin I, Stephen S, Kent WJ, Mattick JS,
Haussler D (2004) Ultraconserved elements in the human
genome. Science 304:1321–1325
Bork P, Copley R (2001) The draft sequences. Filling in the gaps.
Nature 409:818–820
delaCalle-MustienesE,FeijooCG,ManzanaresM,TenaJJ,Rodriguez-
Seguel E, Letizia A, Allende ML, Gomez-Skarmeta JL (2005) A
functionalsurveyoftheenhanceractivityofconservednon-coding
sequences from vertebrate Iroquois cluster gene deserts. Genome
Res 15:1061–1072
Delsuc F, Brinkmann H, Chourrout D, Philippe H (2006) Tunicates
and not cephalochordates are the closest living relatives of
vertebrates. Nature 439:965–968
Dermitzakis ET, Reymond A, Antonarakis SE (2005) Conserved non-
genicsequences—an unexpectedfeatureofmammaliangenomes.
Nat Rev Genet 6:151–157
Dunn CW, Hejnol A, Matus DQ, Pang K, Browne WE, Smith SA,
Seaver E, Rouse GW, Obst M, Edgecombe GD et al (2008)
Broad phylogenomic sampling improves resolution of the animal
tree of life. Nature 452:745–749
Ferguson-Smith MA, Trifonov V (2007) Mammalian karyotype
evolution. Nat Rev Genet 8:950–962
Garland TJ, Ives AR (2000) Using the past to predict the present:
confidence intervals for regression equations in phylogenetic
comparative methods. Am Nat 155:346–364
Gregory RT (2005) Synergy between sequence and size in the study
of genomes. Nat Rev Genet 6:699–708
Gregory RT, DeSalle R (2005) Comparative genomics in prokaryotes.
In: Gregory RT (ed) The evolution of the genome. Elsevier, San
Diego, CA, pp 585–675
Guindon S, Gascuel O (2003) PhyML – A simple, fast, and accurate
algorithm to estimate large phylogenies by maximum likelihood.
Syst Biol 52(5):696–704
Itoh T, Toyoda A, Taylor TD, Sakaki Y, Hattori M (2005)
Identification of large ancient duplications associated with human
gene deserts. Nat Genet 37:1041–1043
Karolchik D, Baertsch R, Diekhans M, Furey TS, Hinrichs A, Lu YT,
Roskin KM, Schwartz M, Sugnet CW, Thomas DJ et al (2003)
The UCSC Genome Browser Database. Nucleic Acids Res
31:51–54
Keane TM, Creevey CJ, Pentony MM, Naughton TJ, McInerney JO
(2006) Assessment of methods for amino acid matrix selection
and their use on empirical data shows that ad hoc assumptions
for choice of matrix are not justified. BMC Evol Biol 6:29
Kidwell MG (2002) Transposable elements and the evolution of
genome size in eukaryotes. Genetica 115:49–63
Kissinger JC, Brunk BP, Crabtree J, Fraunholz MJ, Gajria B, Milgram
AJ, Pearson DS, Schug J, Bahl A, Diskin SJ et al (2002) The
Plasmodium genome database. Nature 419:490–492
Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, Baldwin J,
Devon K, Dewar K, Doyle M, FitzHugh W et al (2001) Initial
sequencingandanalysisofthehumangenome.Nature409:860–921
Lartillot N, Brinkmann H, Philippe H (2007) Suppression of long-
branch attraction artefacts in the animal phylogeny using a site-
heterogeneous model. BMC Evol Biol 7(Suppl. 1):S4
Lindblad-Toh K, Wade CM, Mikkelsen TS, Karlsson EK, Jaffe DB,
Kamal M, Clamp M, Chang JL, Kulbokas EJ 3rd, Zody MC et al
(2005) Genome sequence, comparative analysis and haplotype
structure of the domestic dog. Nature 438:803–819
Lynch M (2007) The origins of genome architecture. Sinauer,
Sunderland, MA
Lynch M, Conery JS (2003) The origins of genome complexity.
Science 302:1401–1404
Maddison WP, Maddison DR (2004) Mesquite: a modular system for
evolutionary analysis. www.mesquiteproject.org
Messing J, Bharti AK, Karlowski WM, Gundlach H, Kim HR, Yu Y,
Wei F, Fuks G, Soderlund CA, Mayer KF et al (2004) Sequence
composition and genome organization of maize. Proc Natl Acad
Sci U S A 101:14349–14354
Mikkelsen TS, Wakefield MJ, Aken B, Amemiya CT, Chang JL,
Duke S, Garber M, Gentles AJ, Goodstadt L, Heger A et al
(2007) Genome of the marsupial Monodelphis domestica reveals
innovation in non-coding sequences. Nature 447:167–177
Nelson CE, Hersh BM, Carroll SB (2004) The regulatory content of
intergenic DNA shapes genome architecture. Genome Biol 5:R25
Nobrega MA, Ovcharenko I, Afzal V, Rubin EM (2003) Scanning
human gene deserts for long-range enhancers. Science 302:413
Nobrega MA, Zhu Y, Plajzer-Frick I, Afzal V, Rubin EM (2004)
Megabase deletions of gene deserts result in viable mice. Nature
431:988–993
Ohno S (1985) Dispensable genes. Trends Genet 1:160–164
Ovcharenko I, Loots GG, Nobrega MA, Hardison RC, Miller W,
Stubbs L (2005) Evolution and functional classification of
vertebrate gene deserts. Genome Res 15:137–145
J Mol Evol (2009) 69:207–216 215
123

Page 10

Sandelin A, Bailey P, Bruce S, Engstrom PG, Klos JM, Wasserman
WW, Ericson J, Lenhard B (2004) Arrays of ultraconserved non-
coding regions span the loci of key developmental genes in
vertebrate genomes. BMC Genomics 5:99
Siepel A, Bejerano G, Pedersen JS, Hinrichs AS, Hou M, Rosenbloom
K, Clawson H, Spieth J, Hillier LW, Richards S et al (2005)
Evolutionarily conserved elements in vertebrate, insect, worm,
and yeast genomes. Genome Res 15:1034–1050
Taylor J (2005) Clues to function in gene deserts. Trends Biotechnol
23:269–271
Venter JC, Adams MD, Myers EW, Li PW, Mural RJ, Sutton GG,
Smith HO, Yandell M, Evans CA, Holt RA et al (2001) The
sequence of the human genome. Science 291:1304–1351
Xu L, Chen H, Hu X, Zhang R, Zhang Z, Luo ZW (2006) Average
gene length is highly conserved in prokaryotes and eukaryotes
and diverges only between the two kingdoms. Mol Biol Evol
23:1107–1108
216J Mol Evol (2009) 69:207–216
123