Page 1
and Evolution. All rights reserved. For permissions, please e-mail: [email protected] The Author 2008. Published by Oxford University Press on behalf of the Society for Molecular Biology
1
Convergent evolution of clustering of Iroquois homeobox genes across metazoans
Letter
Manuel Irimia*, Ignacio Maeso*, Jordi Garcia-Fernàndez§
Departament de Genètica, Facultat de Biologia, Universitat de Barcelona,
Av. Diagonal 645, 08028, Barcelona, Spain.
* These authors contributed equally to this work.
§ Corresponding author.
Jordi Garcia-Fernandez
e-mail: [email protected]
Ph: +34 934034437
Fax: +34 934034420
Key words: Amphioxus, Iroquois, Genome evolution, gene cluster, convergent evolution,
synteny conservation
Running Head: Convergent clustering of Iroquois genes
Title length: 78 characters
Abstract length: 164 words
Main body: 14,653 characters
Page requirement: 5.0 printed pages
References: 23
MBE Advance Access published May 9, 2008 by guest on July 1, 2015
http://mbe.oxfordjournals.org/
Dow
nloaded from
Page 2
2
Convergent evolution of clustering of Iroquois homeobox genes across metazoans
Manuel Irimia*, Ignacio Maeso*, Jordi Garcia-Fernàndez§
Departament de Genètica, Facultat de Biologia, Universitat de Barcelona,
Av. Diagonal 645, 08028, Barcelona, Spain.
* These authors contributed equally to this work.
§ To whom correspondence should be addressed. E-mail: [email protected]
Vertebrate and Drosophila Iroquois genes are organized in clusters of three genes
sharing blocks of conserved regulatory sequences. Here, we report a three-gene cluster
in the basal, pre-duplicative chordate amphioxus. Surprisingly, however, the origin of
the amphioxus cluster is independent of those in vertebrates and drosophilids.
Investigation of genomic organization of Iroquois genes in other 17 metazoan genomes
revealed a fourth independent three-gene cluster organization in polychaetes, as well as
additional two- and four-gene clusters in other clades, in one of the most striking
examples of convergence in genomic organization described so far. The recurrent
independent evolution of Iroquois clusters suggests a functional importance of this
organization for these genes, perhaps related to the sharing of regulatory elements.
Consistent with this, comparative analysis of genomic regions flanking the three
amphioxus Irx genes revealed several blocks of sequences, conserved for at least 100
million years. Finally, we discuss the possible causes and implications of the convergent
evolution of this genomic and regulatory organization throughout metazoans.
Restructuring and shuffling of genome architecture is a major source for evolutionary
change; however, little is known about the mechanisms underlying these changes. With the
completion of several metazoan genome sequencing projects, many key features of genome
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 3
3
structure have been discovered; the finding of gene regulatory blocks, conserved synteny, and
gene deserts have given rise to stimulating hypothesis about the origin, conservation and
evolution of genomic structure and function. In this regard, the presence of conserved cis
regulatory elements has been shown to be crucial both in maintaining gene clustering, and in
promoting extensive gene-free regions (Nobrega et al. 2003; Ovcharenko et al. 2005;
Engstrom et al. 2007; Kikuta et al. 2007).
Iroquois genes represent a paradigmatic example of the relationship between gene
regulation and genome organization (de la Calle-Mustienes et al. 2005). Iroquois genes are
homeobox transcription factors of the TALE super-class implicated in key developmental
processes during animal development. In vertebrates, Irx genes regulate proneural genes
(Bellefroid et al. 1998; Gómez-Skarmeta et al. 1998a), and are involved in several other
processes during gastrulation, nervous system regionalization and organ patterning (Briscoe et
al. 2000; Gómez-Skarmeta, de La Calle-Mustienes, and Modolell 2001; Kobayashi et al.
2002). Perhaps the most prominent feature of Irx genes is their functional genomic
organization, consisting of complexes of three clustered genes. In tetrapods, due to the whole
genome duplications occurred at the origin of the vertebrate lineage (Dehal and Boore 2005),
there are six Irx genes grouped into two paralogous genomic clusters: IrxA, containing Irx1,
Irx2, and Irx4, and IrxB, containing Irx3, Irx5, and Irx6 (Peters et al. 2000). Interestingly, the
developmental expression patterns of Irx1 and Irx2 and of Irx3 and Irx5, respectively, are
almost identical, whereas the expression of the third gene of each cluster, Irx4 or Irx6, is
generally more divergent (Bellefroid et al. 1998; Gómez-Skarmeta et al. 1998b; Garriock et
al. 2001; Houweling et al. 2001). These expression patterns can be explained by the
differential distribution of several highly conserved non-coding regions (HCNRs, (Sandelin et
al. 2004; Woolfe et al. 2005)) within the clusters, which drive expression in the common
territories where several Irx genes are expressed (de la Calle-Mustienes et al. 2005). These
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 4
4
HCNRs act as enhancers shared by more than one gene, and provide an explanation as to why
vertebrate Irx genes have remained in cluster (de la Calle-Mustienes et al. 2005).
Interestingly, a parallel situation is observed in Drosophila, where the three Iroquois genes
also share expression domains and specific regulatory elements (Gomez-Skarmeta and
Modolell 2002), although the vertebrate and fruit fly clusters originated independently (Peters
et al. 2000; de la Calle-Mustienes et al. 2005).
Here, we study the genomic organization of Iroquois genes in the basal, pre-
duplicative chordate amphioxus Branchiostoma floridae. As in vertebrates, we found a cluster
of three genes (BfIrxA, BfIrxB and BfIrxC). To explore the evolution of these clusters, we
performed gene tree analyses using phylogenetic methods. Interestingly, the three B. floridae
genes form a clade (Figure SM1), indicating that these three genes represent duplications of a
single ancestral gene after the divergence of cephalochordates from other chordates. The
clusters’ independent origins are also underscored by the different strand orientation of the
genes (two genes 3’-to-3’ in Branchiostoma, two genes 5’-to-5’ in vertebrates, Figure 1). In
addition, a search using RT-PCR for Irx genes in a sister amphioxus species, Branchiostoma
lanceolatum, split from B. floridae around 100 million years ago (Cañestro et al. 2002;
Nohara, Nishida, and Nishikawa 2005; Kon et al. 2007), yield orthologs for the three B.
floridae Irx genes. All three amphioxus genes are expressed in a similar temporal manner
during whole embryonic development, with a peak of expression from neurula to 1-gill-slit
larva stages, as indicated by RT-PCR analysis (data not shown).
Genomic analysis of other 17 available metazoan genomes, including species from
placozoans, cnidarians and the three main bilaterian clades (Ecdysozoans, Lophotrochozoans
and Deuterostomes) revealed further surprises. Strikingly, we found a fourth independent
three-gene cluster of Iroquois genes in the polychaete Capitella capitata. Another cluster, of
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 5
5
four genes, was found in the mollusk Lottia gigantea, and two other independent clusters,
consisting of two genes each, were found in each of the two tunicates species examined
(Ciona savigni and Ciona intestinalis, as previously reported (Wada et al. 2003)) (Figure 1).
In each case, genes within a lineage show greater similarity to each other than to genes from
other lineages, indicating recurrent gene duplication leading to these clusters (Figure SM1).
Considering that the basal metazoans groups (placozoans and cnidarians) and the non-
chordate deuterostomes harbor a single Iroquois gene (Figure 1), the observed pattern implies
at least 13 tandem gene duplications of Iroquois genes in six independent lineages, with
further convergence on exactly three-gene clusters in four deeply diverged metazoan
groups. This is, to our knowledge, the most remarkable example of convergence in genomic
organization described so far.
What explains this recurrent convergent evolution? In vertebrate models, the presence
of shared HCNRs acting as regulatory elements, particularly the ultraconserved regions
(UCRs), may act as a constraint to maintain the cluster integrity (de la Calle-Mustienes et al.
2005). Does the amphioxus cluster show similar organization? We searched for the elements
described within the vertebrate Irx cluster in amphioxus, but failed to find a clear match,
consistent with independent evolutionary origins of the Branchiostoma and vertebrate
clusters. However, comparison of the genomic regions surrounding the three B. floridae Irx
genes revealed lineage-specific conserved regions. We divided the amphioxus cluster into
three regions, each containing one of the genes and surrounding non-coding sequences (see
Supplementary Methods Online). Crossed VISTA (Frazer et al. 2004) analysis of these
regions revealed several repeated blocks with high sequence similarity (Figure 2a). Most
blocks are present in two copies, but some are present in three copies. Three-copy sequences
lie nearby the coding sequences (one for each gene; Figure 2b, yellow boxes), consistent with
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 6
6
being part of minimal promoters, UTRs, or other proximal cis elements. Two-copy blocks are
present in opposite orientation, in agreement with the opposite transcriptional orientations of
the genes, and are located further away from the coding sequences and may thus be able to
function over longer distances (and so to influence expression of all three genes; Figure 2b,
turquoise and red boxes). These two-copy elements likely were initially doubly duplicated
along with the coding sequences, but have since been differentially lost (Figure 2c).
Importantly, we have also identified and cloned these blocks in B. lanceolatum, which
indicates that they have been conserved for around 100 mya (Cañestro et al. 2002; Nohara,
Nishida, and Nishikawa 2005; Kon et al. 2007), strongly suggesting a functional role for these
elements.
The finding of potential regulatory elements that are present in fewer than three copies
is consistent with a role for these elements in stabilizing the Irx cluster in Branchiostoma as in
other metazoans (Gomez-Skarmeta and Modolell 2002; de la Calle-Mustienes et al. 2005).
Whereas the initial duplicated genomic region might have contained all the necessary
regulatory sequences, loss of some of these elements might have rendered the genes
dependent on proximity to the other copies, ensuring the continued linkage of the paralogs in
a cluster. At the same time, the presence of multiple Irx genes would allow for the emergence
of copy-specific regulatory elements and patterns, aiding the emergence of new
developmental programs.
Related to the presence of non-coding regulatory sequences, genes with complex
regulation usually show large surrounding genomic regions devoid of genes (gene deserts)
(Nelson, Hersh, and Carroll 2004). Similarly, clusters of highly regulated genes usually have
large intergenic regions between cluster members, as is the case of the Irx clusters in
vertebrates (de la Calle-Mustienes et al. 2005). The amphioxus Irx cluster also shows very
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 7
7
large intergenic regions within the cluster, much larger than the distances between the genes
in the proximity and than the average genomic intergenic distance (Table 1). A similar pattern
is observed for clusters of all species and in the case of the surrounding regions of the species
with a single Irx gene, with the exception of the placozoans and cnidarians (Table 1 and
Supplementary file 1).
Interestingly, the study of the Irx surrounding regions across metazoans yielded
another surprising observation. Amphioxus Irx cluster is flanked directly upstream by an
ortholog of the Drosophila CG10632 locus and downstream by the Carbonic anhydraseVIII
(CAVIII) gene. This synteny is conserved to sea urchin (where only a single Irx is present),
making sea urchin's organization highly resembling of the hypothesized ancestral Irx
organization in the amphioxus lineage (Figure 2C). (Unfortunately, we were not able to
identify any clearly conserved non-coding sequence between amphioxus and sea urchin using
VISTA analysis). Strikingly, CG10632 orthologs are also immediately 5' of the Irx cluster in
all studied insects and in the crustacean Daphnia pulex (black arrowheads in Figure 1).
Moreover, in lophotrochozoans, the CG10632 ortholog is within the Irx cluster, shedding light
on the events leading to lophotrochozoan's Irx cluster formation and further supporting the
independent evolution of this cluster. In the case of CAVIII, although the synteny is not
conserved to protostomes, this gene is flanking upstream the single Irx gene in placozoans
(white arrowhead in Figure 1).
These data strongly support that the linkage between CG10632, CAVIII and Iroquois
genes, which has been retained in amphioxus and sea urchin, is ancestral at least to
bilaterians. In the case of CG10632, this linkage has been maintained in members of all major
metazoans groups, encompassing more than 600 de million years of evolution, in one of the
most striking examples of conserved synteny between two phylogenetically unrelated genes.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 8
8
For this reason, we propose for this gene the name Sosondowah (sowah), after the
mythological hunter tied to her doorpost by an Iroquois goddess.
In summary, the results presented here show that the organization of Iroquois genes in
gene complexes has evolved independently several times in metazoan evolution, especially
generating clusters of three genes. The number of convergent clusterization events thus seems
to exceed what would be expected merely by chance. If so, it is tempting to speculate that
tandem duplication and neighborhood organization for Iroquois genes is fixed in evolution at
a high rate, maybe related to the functional importance of shared regulatory elements.
Intriguingly, our results would suggest that gene tandem duplication is more widespread than
previously thought, being only fixed at high rates in cases in which a cluster organization bear
a functional relevance, as it seems the case for the Iroquois complexes. An alternative
explanation is that some genomic regions are more prone to undergo tandem duplication;
however, there is currently no evidence that specific groups of genes or types of sequences are
more likely to duplicate than others. Whatever the case, the study of Iroquois gene
organization will help to understand how and why functional architectures evolve in
metazoans genomes. Understanding the causes and mechanisms involved in the evolution of
these genes will help to understand the flexibility, constraints, and evolvability of genome
organization.
Acknowledgements
We thank Scott W. Roy for critical reading of the manuscript and extremely helpful
comments and suggestions and Jim Langeland for kindly providing the B. floridae cDNA
library. We particularly thank an anonymous referee for insightful suggestions. This work
was funded by grant BFU2005-00252 from the Ministerio de Educación y Ciencia (MEC),
Spain. MI holds FPI and IM FPU fellowships (MEC).
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 9
9
References
Bellefroid, E. J., A. Kobbe, P. Gruss, T. Pieler, J. B. Gurdon, and N. Papalopulu. 1998. Xiro3
encodes a Xenopus homolog of the Drosophila Iroquois genes and functions in neural
specification. EMBO J 17:191-203.
Briscoe, J., A. Pierani, T. M. Jessell, and J. Ericson. 2000. A homeodomain protein code
specifies progenitor cell identity and neuronal fate in the ventral neural tube. Cell
101:435-445.
Cañestro, C., R. Albalat, L. Hjelmqvist, L. Godoy, H. Jarnvall, and R. Gonzalez-Duarte. 2002.
Ascidian and Amphioxus Adh Genes Correlate Functional and Molecular Features of
the ADH Family Expansion During Vertebrate Evolution. Journal of Molecular
Evolution 54:81-89.
de la Calle-Mustienes, E., C. G. Feijoo, M. Manzanares, J. J. Tena, E. Rodriguez-Seguel, A.
Letizia, M. L. Allende, and J. L. Gomez-Skarmeta. 2005. A functional survey of the
enhancer activity of conserved non-coding sequences from vertebrate Iroquois cluster
gene deserts. Genome Res. 15:1061-1072.
Dehal, P., and J. L. Boore. 2005. Two rounds of whole genome duplication in the ancestral
vertebrate. PLoS Biol 3:e314.
Engstrom, P. G., S. J. Ho Sui, O. Drivenes, T. S. Becker, and B. Lenhard. 2007. Genomic
regulatory blocks underlie extensive microsynteny conservation in insects. Genome
Res. 17:1898-1908.
Frazer, K. A., L. Pachter, A. Poliakov, E. M. Rubin, and I. Dubchak. 2004. VISTA:
computational tools for comparative genomics. Nucleic Acids Res. 32:W273-279.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 10
10
Garriock, R., S. Vokes, E. Small, R. Larson, and P. Krieg. 2001. Developmental expression of
the Xenopus Iroquois-family homeobox genes, Irx4 and Irx5. Development Genes and
Evolution 211:257-260.
Gómez-Skarmeta, J., E. de La Calle-Mustienes, and J. Modolell. 2001. The Wnt-activated
Xiro1 gene encodes a repressor that is essential for neural development and
downregulates Bmp4. Development 128:551-560.
Gómez-Skarmeta, J. L., A. Glavic, E. de la Calle-Mustienes, J. Modolell, and R. Mayor.
1998a. Xiro, a Xenopus homolog of the Drosophila Iroquois complex genes, controls
development at the neural plate. EMBO J 17:181-190.
Gómez-Skarmeta, J. L., A. Glavic, E. de la Calle-Mustienes, J. Modolell, and R. Mayor.
1998b. Xiro, a Xenopus homolog of the Drosophila Iroquois complex genes, controls
development at the neural plate. EMBO J. 17:181-190.
Gomez-Skarmeta, J. L., and J. Modolell. 2002. Iroquois genes: genomic organization and
function in vertebrate neural development. Current Opinion in Genetics &
Development 12:403-408.
Houweling, A. C., R. Dildrop, T. Peters, J. Mummenhoff, A. F. M. Moorman, U. Ruther, and
V. M. Christoffels. 2001. Gene and cluster-specific expression of the Iroquois family
members during mouse development. Mechanisms of Development 107:169-174.
Kikuta, H., D. Fredman, S. Rinkwitz, B. Lenhard, and T. Becker. 2007. Retroviral enhancer
detection insertions in zebrafish combined with comparative genomics reveal genomic
regulatory blocks - a fundamental feature of vertebrate genomes. Genome Biology
8:S4.
Kobayashi, D., M. Kobayashi, K. Matsumoto, T. Ogura, M. Nakafuku, and K. Shimamura.
2002. Early subdivisions in the neural plate define distinct competence for inductive
signals. Development 129:83-93.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 11
11
Kon, T., M. Nohara, Y. Yamanoue, Y. Fujiwara, M. Nishida, and T. Nishikawa. 2007.
Phylogenetic position of a whale-fall lancelet (Cephalochordata) inferred from whole
mitochondrial genome sequences. BMC Evol Biol 7:127.
Nelson, C. E., B. M. Hersh, and S. B. Carroll. 2004. The regulatory content of intergenic
DNA shapes genome architecture. Genome Biol. 5:R25.
Nobrega, M. A., I. Ovcharenko, V. Afzal, and E. M. Rubin. 2003. Scanning Human Gene
Deserts for Long-Range Enhancers. Science 302:413-.
Nohara, M., M. Nishida, and T. Nishikawa. 2005. New complete mitochondrial DNA
sequence of the lancelet Branchiostoma lanceolatum (Cephalochordata) and the
identity of this species' sequences. Zoolog. Sci. 22:671-674.
Ovcharenko, I., G. G. Loots, M. A. Nobrega, R. C. Hardison, W. Miller, and L. Stubbs. 2005.
Evolution and functional classification of vertebrate gene deserts. Genome Res.
15:137-145.
Peters, T., R. Dildrop, K. Ausmeier, and U. Ruther. 2000. Organization of Mouse Iroquois
Homeobox Genes in Two Clusters Suggests a Conserved Regulation and Function in
Vertebrate Development. Genome Res. 10:1453-1462.
Sandelin, A., P. Bailey, S. Bruce, P. Engstrom, J. Klos, W. Wasserman, J. Ericson, and B.
Lenhard. 2004. Arrays of ultraconserved non-coding regions span the loci of key
developmental genes in vertebrate genomes. BMC Genomics 5:99.
Wada, S., M. Tokuoka, E. Shoguchi et al. 2003. A genomewide survey of developmentally
relevant genes in Ciona intestinalis. II. Genes for homeobox transcription factors. Dev
Genes Evol 213:222-234.
Woolfe, A., M. Goodson, D. K. Goode et al. 2005. Highly Conserved Non-Coding Sequences
Are Associated with Vertebrate Development. PLoS Biology 3:e7.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 12
12
Figure Legends
Figure 1 – Repeated clusterization of Iroquois genes and conserved synteny in
metazoans.
Iroquois genes are represented by arrows, showing transcriptional orientation. Orthologs of
the Drosophila melanogaster locus CG10625 (or sowah) are represented by black arrowheads
and orthologs of the CAVIII by white arrowheads, both showing transcriptional orientation.
Question marks in Saccoglossus and Lottia indicate that the genomic regions where these
genes would be expected are not available. The phylogenetic relationship and evolutionary
history of the genes on the most parsimonious evolutionary scenario is illustrated. First, gene
and genome duplication events are indicated by red and blue stars, respectively. Second,
colors are used to identify genes that are most phylogenetically related.
Figure 2 - Internal cluster organization and evolutionary origin of B. floridae Irx genes.
A) VISTA plot of the alignments between each of the three Irx genes and their respective
surrounding non-coding regions. Colored peaks (blue, coding; pink, non-coding) indicate
regions of at least 100 bp and 70% similarity. Colored bars indicate the peaks depicted in B.
B) Schematic organization of the conserved sequence blocks within the B. floridae Irx cluster.
Red and turquoise boxes enclose large putative ancestral regions containing several conserved
blocks. Black block arrows indicate the coding sequences of the three Irx genes, showing
transcriptional orientation. Vertical bars of different colors represent the different conserved
repeated blocks, as indicated in A. Vertical discontinuous lines delimit the three regions that
were used in the VISTA analysis. C) Putative evolutionary scenario for the origin of the B.
floridae Irx cluster. First, an ancestral IrxABC gene and some of the surrounding conserved
non-coding sequences (those directly upstream) duplicated in tandem, giving rise to the IrxB
gene and an ancestral IrxAC gene; IrxB subsequently lost and rearranged some of these
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 13
13
conserved blocks. IrxAC and the 5’ and 3’ surrounding conserved blocks then duplicated and
reverse orientation, originating IrxA and IrxC. Finally, IrxA lost some of the upstream
conserved blocks.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 14
14
Table 1 - Intergenetic distances between Irx and neighbor genes across metazoans, and average and median genomic intergenic
distances.
Species Median
intergenic distance
Average. intergenic distance
Average intergenic
distance (5’)
Distance to flanking gene (5')
(to sowah) Distance
between Irx genes
(to CAVIII) Distance to
flanking gene (3')
Average intergenic
distance (3’) B. floridae 3933 8608.12 3781.1 29739 14619 56689 31310 12064 9564.2 H. sapiens cluster1-2-4 18205 48313.13 114827.2 1432980 -- 854199.5 -- 61384 50121.2 H. sapiens cluster3-5-6 18205 48313.13 58819.2 168831 -- 517403.5 -- 148416 49008 M. musculus cluster1-2-4 17660 56656.31 143579.9 403034 -- 645930 -- 48228 27392 M. musculus cluster3-5-6 17660 56656.31 77501.8 108389 -- 434684 -- 35636 22117.9 C. intestinalis clusterA-B 2253 5693.76 2615.6 8795 -- 32617 -- 14569 2659.6 C. intestinalis clusterC-D 2253 5693.76 2868.1 12723 -- 33565 -- 3381 3554.6 S. purpuratus 7233 17892.95 n.d. >113507 49585 -- 127140 >10059 n.d. C. capitata 3571 6043.89 4461* 4036 7484 31870.5 -- 9991 7023.5* L. gigantea 3869 8969.06 19244.6 53488 ? 128605 -- 29569 8215.2 D. melanogaster 798 5519.59 343.5 1994 8152 43024 -- 23216 410.7 A. gambiae 3394 15483.46 4006.1 2615 73514 57339 -- 150350 7064.4 A. mellifera 2071 10702.25 >11516.5 1060 9679 16254 -- 346396 >2758.2 N. vitripennis 2952 12688.24 2812 2766 8427 54098 -- 114367 19416.2 T. castaneum 1618 7104.92 53228.9 1051 322 63137 -- 176682 17106.5 D. pulex 1493 3610.56 4492.9 157 18300 27649 -- 114986 3998.7 N. vectensis 3759 6722.02 13467.1 10482 -- -- -- 28771 14948.7 T. adhaerens 2806 5446.36 4647.4 2931 -- -- 28044** 7933 10292.1
* Average intergenic distance at 5' and 3' calculated with less than 10 neighbor genes, due to incompleteness of the assembly. ** Trichoplax adhaerens Irx has been likely inverted. Thus, the CAVIII ortholog is at 5' of Irx. n.d. Not data available.
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 15
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from
Page 16
by guest on July 1, 2015http://m
be.oxfordjournals.org/D
ownloaded from