Genome Structure of the Legume, Lotus japonicus.

Genome Structure of the Legume, Lotus japonicus
Shusei SATO1, Yasukazu NAKAMURA1, Takakazu KANEKO1, Erika ASAMIZU1, Tomohiko KATO1, Mitsuteru NAKAO1,
Shigemi SASAMOTO1, Akiko WATANABE1, Akiko ONO1, Kumiko KAWASHIMA1, Tsunakazu FUJISHIRO1,
Midori KATOH1, Mitsuyo KOHARA1, Yoshie KISHIDA1, Chiharu MINAMI1, Shinobu NAKAYAMA1, Naomi NAKAZAKI1,
Yoshimi SHIMIZU1, Sayaka SHINPO1, Chika TAKAHASHI1, Tsuyuko WADA1, Manabu YAMADA1, Nobuko OHMIDO2,
Makoto HAYASHI3, Kiichi FUKUI3, Tomoya BABA4, Tomoko NAKAMICHI5, Hirotada MORI5, and Satoshi TABATA1,*
Kazusa DNA Research Institute, 2-6-7 Kazusa-kamatari, Kisarazu, Chiba 292-0818, Japan1; Graduate School of
Human Development and Environment, Kobe University, Kobe 657-8501, Japan2; Department of Biotechnology,
Graduate School of Engineering, Osaka University 2-1 Yamadaoka, Suita 565-0871, Osaka, Japan3; Institute of
Advanced Biosciences, Keio University, Tsuruoka, Yamagata 997-0017, Japan4 and Graduate School of Biological
Sciences, Nara Institute of Science and Technology, 8916-5 Takayama, Ikoma, Nara 630-0192, Japan5
(Received 1 April 2008; accepted on 18 April 2008; published online 28 May 2008)
Abstract
The legume Lotus japonicus has been widely used as a model system to investigate the genetic back-
ground of legume-specific phenomena such as symbiotic nitrogen fixation. Here, we report structural fea-
tures of the L. japonicus genome. The 315.1-Mb sequences determined in this and previous studies
correspond to 67% of the genome (472 Mb), and are likely to cover 91.3% of the gene space. Linkage
mapping anchored 130-Mb sequences onto the six linkage groups. A total of 10 951 complete and
19 848 partial structures of protein-encoding genes were assigned to the genome. Comparative analysis
of these genes revealed the expansion of several functional domains and gene families that are character-
istic of L. japonicus. Synteny analysis detected traces of whole-genome duplication and the presence of
synteny blocks with other plant genomes to various degrees. This study provides the first opportunity
to look into the complex and unique genetic system of legumes.
Key words: Lotus japonicus; genome structure; Fabaceae; comparative analysis
1. Introduction
Fabaceae is the third largest family of flowering
plants, comprising 650 genera and 18 000 species
with a variety of characteristics; many of which have
long been targets of breeding because of their agro-
nomic and industrial importance. Among them, a
few species have been chosen as ‘model legumes’ for
use in genetic and physiological studies. Lotus japoni-
cus is a typical model legume with the characteristics
of a short life cycle (2–3 months), self-fertility, and
a relatively simple genome architecture of diploidy
(n ¼ 6), i.e. small in size, 472 Mb. Mutants in
various biological phenomena specific to legumes
such as symbiotic nitrogen fixation, and those
common to flowering plants such as flower morpho-
genesis, have been characterized and the genes
responsible have been isolated and further studied.
The availability of the Agrobacterium-mediated DNA
transformation system and genomic resources includ-
ing a large number of expressed sequences tag
(EST)/cDNA clones,1 high-density genetic linkage
maps,2–4 and partial genome sequences5–9 has
played an essential role in this process.
The whole-genome sequences of two plant species,
Arabidopsis thaliana (Cruciferae) and Oryza sativa
(Poaceae), have drastically accelerated research into
their genetic systems by providing investigators with
both gene sequences and positional information.
Edited by Katsumi Isono
* To whom correspondence should be addressed. Tel. þ81 438-
52-3933. Fax. þ81 438-52-3934. E-mail: tabata@kazusa.or.jp
# The Author 2008. Kazusa DNA Research Institute.
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the
open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal
and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently
reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use,
please contact journals.permissions@oxfordjournals.org
DNA RESEARCH 15, 227–239, (2008) doi:10.1093/dnares/dsn008

However, the process of high-accuracy sequencing is
expensive and labor intensive since physical maps are
created and a large number of individual BAC clones
are used as templates. In contrast, draft sequencing
according to the whole-genome shotgun approach in
combinationwith BACend sequencing generates infor-
mation about the general genome structure at a fairly
reasonable cost, but the resulting data are rather frag-
mentary and need an additional effort to be connected
with physical/genetic maps. In this study, we aimed to
analyze the genome structure of L. japonicus to reveal
gene and genome features that are characteristic of
legume plants. For this purpose, we adopted various
established technologies including genomic library
construction, DNA sequencing, fluorescent in situ
hybridization (FISH), genetic mapping and bioinfor-
matics in such a way that the advantages of each tech-
nology were combined in a cost-effective manner.
Here, we report the first whole-genome structure of
the legume, L. japonicus, its characteristic genome fea-
tures, and a variety of information and material
resources that were developed during this study.
2. Material and methods
2.1. Plant and DNA materials
Lotus japonicus accession Miyakojima MG-20 was
provided by Masayoshi Kawaguchi, University of
Tokyo.10 Transformation-competent artificial chromo-
some (TAC) genomic libraries were constructed
according to standard methods as described pre-
viously.5 BAC genomic libraries were constructed
using the genomic DNA of L. japonicus accession
MG-20 partially digested with either Sau3AI or EcoRI
and pBeloBAC as a cloning vector. The average insert
size of these libraries was 104 kb for the Sau3AI
library and 101 and 88 kb for two independent pre-
parations of EcoRI libraries. Both libraries covered the
haploid genome 8.4 times in total.
The seeds and genomic libraries can be obtained
from LegumeBase supported by the National
BioResource Project (http://www.legumebase.agr.
miyazaki-u.ac.jp/index.jsp).
2.2. Genome sequencing and assembly
Two types of sequencing approaches were com-
bined to sequence the L. japonicus genome: clone-
by-clone sequencing and shotgun sequencing of
selected regions of the genome.
TAC/BAC clones were selected from the genomic
libraries as seed points using the sequence infor-
mation from ESTs and cDNA markers from L. japonicus
and other legumes. The nucleotide sequence of each
clone was determined according to the shotgun strat-
egy with three to five times redundancy. A total of
1909 TAC/BAC clones, those newly sequenced in
this study and those that had been sequenced
previously,5–9 were assembled into 954 scaffolds
using the Paracel Genome Assembler (PGA;
version 2.6.2, Paracel Co., 2002), followed by
manual TAC/BAC end-pair scaffolding, resulting in
high-quality genomic sequence (HGS) contigs.
In parallel, shotgun sequencing of a selected TAC
mixture (STM) enriched in gene spaces and a whole
genomic DNA from which highly repetitive and orga-
nelle genomic sequences were subtracted (selected
genomic regions, SGRs) was carried out. The TAC
clones, neither end sequence of which hit repetitive
or organelle genomic sequences in the L. japonicus
genome, were selected from the libraries, pooled,
and subjected to shotgun sequencing. For the SGRs,
a genomic library with an average insert size of
2.5 kb was generated using pBluescript SK2 as the
cloning vector. For subtraction, polymerase chain
reaction (PCR)-amplified fragments of LjTR1 were bio-
tinylated using Biotin-High Prime (Roche, Basel,
Switzerland) and used as a driver in subtractive
hybridization with the WGS library. The WGS library
was single-stranded prior to hybridization by com-
bined action of gene II and exonuclease III. Hybrids
were removed using Dynabeads M-280 Streptavidin
(Invitrogen, Carlsbad, CA, USA) and the remaining
single-stranded WGS library was double-stranded
using Klenow fragments (Takara Bio, Japan) and trans-
formed into host E. coli ElectroTen-Blue (Agilent
Technologies, Santa Clara, CA, USA).
A total of 808 816 reads from STM generated from
4603 TAC inserts and 847 513 SGR reads were
assembled into a set of 109 986 contigs, 147 805
446 bp in length (selected genome assembly, SGA)
by the Arachne assembler, version 2.01.11 The SGA
sequences were then subjected to assemble with the
HGS, and finally, a total of 110 940 supercontigs
with a total coverage of 315 073 275 tentative
genomic sequence (TGS) bases were obtained.
2.3. Linkage mapping
Two types of PCR-based DNA marker, SSLP and
dCAPS,were generatedusing the sequence information
in order to locate sequence contigs on the genetic
linkage map, as described previously.5 The analysis of
segregation data for SSR and dCAPS markers and
linkage map integration were carried out using the F2
mapping population of accessions Miyakojima MG-20
and Gifu B-129, which were previously used for con-
struction of the fine genetic linkage map.2
2.4. Fluorescent in situ hybridization
The 26S rDNA, 5S rDNA, and TAC clones were used
as probes for FISH analysis. The 26S and the 5S rDNA
228 Genome analysis of Lotus japonicus [Vol. 15,

probes were produced by PCR using primer pairs that
were designed based on the rRNA and 5S RNA gene
sequences in the L. japonicus genome.
The FISH analysis using rRNA genes and TAC clones
was performed on well-prepared chromosome
spreads according to the method described pre-
viously.12 The preparations were observed under a flu-
orescence microscope (OLYMPUS BX50) equipped
with a sensitive cooled CCD camera (PXL1400), and
the pro-metaphase chromosome spreads with clear
patterns were photographed using blue or green
light excitation and emission filters. Captured images
were digitally stored in a computer and analyzed
using CHIAS3 imaging software.13
2.5. Repetitive sequences
Repetitive elements in the TGS were identified by
comparing all of the contig units, HGS and those
produced from the STM and SGRs, each using
BLASTN14,15 and processing the outputs using the
RECON program.16 A total of 214 consensus
sequences of repetitive elements that appeared at
least 20 times were identified. The consensus
sequences of these elements were subjected to a simi-
larity search against known repeat elements in the
RepBase (http://www.girinst.org/). For the consensus
sequences with features of Class I or II transposable
elements (TEs), full-length candidate sequences were
identified by comparing 10 kb upstream and down-
stream of the corresponding genomic regions to find
long terminal repeats or terminal inverted repeats.
For unclassified consensus sequences, the longest
representative sequences were selected by comparing
the corresponding genome sequences using the
CLUSTALW multiple alignment program. Full-length
elements of TE and representative sequences of
unclassified repeats were collected into a repeat
sequence library along with the previously reported
TE sequences17 and used as references for
RepeatMasker (www.repeatmasker.org) analysis to
delineate the occurrence of these elements in the
total TGS and TAC/BAC end sequences.
2.6. Assignment of RNA-encoding genes
In order to identify the potential RNA-encoding
genes, a structural RNA sequence library was extracted
from GenBank,18 and searched for homologous
sequences in the L. japonicus genome with the use of
the BLASTN function in BLAST. Transfer RNA genes
were predicted using tRNAscan-SE, version 1.4,19 in a
eukaryotic mode with default parameters. Genes for
small nucleolar RNAs (snoRNAs) and7S large nucleolar
RNAs (LRNAs) were predicted using SnoScan20 with
a yeast model and srpSCAN,21 respectively. A total of
93 genes for small nuclear RNAs including those for
21 U1, 19 U2, 7 U4, 23 U5, 21 U6, and 2 U12 were
identified based on their similarity to known genes.
Candidates formiRNA precursorswere identifiedbya
search using the Arabidopsis mature miRNA sequences,
whichwere retrieved fromthemiRNARegistrydatabase
(http://www.sanger.ac.uk/Software/Rfam/mirna/).22
The L. japonicus genome and EST sequences were
searched for sequence patterns that are characteristic
of miRNA using the FUZZNUC program that is available
at EMBOSS (http://emboss.sourceforge.net/) with the
permission of a maximum of two nucleotide mis-
matches. A stem-loop structure was searched within a
distance of 500 nucleotides from the coding region of
the mature miRNA based on the finding that the
longest Arabidopsis thaliana precursor sequences ident-
ified to date are411nucleotides long.23 Target genes of
the miRNA candidates were searched in the coding
sequences (CDS) and 30 untranslated regions (UTRs)
of the presumptive genes assigned in the TGS.
2.7. Assignment of protein-encoding genes
Tentative genomic sequence was subjected to gene
prediction and modeling by the Kazusa Annotation
PipelinE for Lotus japonicus (KAPSEL).5 The KAPSEL
employs ab initio gene-finding software and similarity
searches in order to generate the elements for gene
model production. The ab initio gene-finding software
used in the pipeline includes GeneMark.hmm,24
Genscan25 and Grail26 using the A. thaliana-trained
matrix. Splice-site candidates were deduced by
NetGene227 and SplicePredictor.28 The similarity
searches to detect potential protein-coding exons
were performed using the BLASTX function of BLAST
against the UniProtKB database.29 The assigned exon
candidates were extracted from the original sequence
library, then mapped on the TGS more precisely using
the dps and nap programs in the program suite of the
analysis and annotation tool (AAT) package.30
Similarity searches of transcript sequences were
performed by aligning the TGS against the Gene
Indices31 for legume species including L. japonicus,
M. truncatula and Glycine max. The assigned transcript
sequences were mapped on the TGS using the dds and
gap2 programs in AAT to confirm working models
of protein-encoding genes. As a result of the auto-
mated annotation process, a total of 19 848 partial
and 10 951 complete models were assigned as
protein-encoding genes in the TGS, except for those
related to TEs. The 76.4-Mb sequences in the HGS
were edited and annotated manually to ensure
high-quality gene prediction.
The genes thus assignedweredenotedby IDswith the
clone (LjT**** for TACs and LjB**** for BACs) or contig
(CM****) names followed by sequential numbers from
one end to another. Of these, manually annotated
No. 4] S. Sato et al. 229

genes on the HGS were followed by “.nc”, and
others were followed by “.nd”. The genes assigned on
the SGA sequences were denoted by IDs with the
assemble consensus names (LjSGA_****) followed by
sequential numbers from one end to another in the
insert.
A global alignment of the genome sequences and
ESTs was performed using the NEEDLE program32,33
that is provided at the EMBOSS site (http://emboss.
sourceforge.net/). To identify a possible TATA box-
like motif for recognition by RNA polymerase II, a
search against the plant cis-acting regulatory DNA
elements (PLACE) database34 (http://www.dna.affrc.
go.jp/PLACE/) was carried out.
2.8. Similarity and domain searches
Functional annotation for deduced L. japonicus pro-
teins was performed by a similarity search against the
genes of known function and a domain analysis. The
similarity search was performed using the gapped
BLASTP function of BLAST against the UniProtKB data-
base and protein-encoding genes deduced in A. thali-
ana,35 rice,36 Populus trichocarpa,37 and grapevine.38
For the analysis of gene families and functional
domains, the predicted proteome was searched
against InterPro.39 InterPro annotations were assigned
to each functional motif and domain, and then the
annotation information was translated into GO func-
tional descriptions.40 GO descriptions were grouped
into the GOslim categories of molecular function
and biological process. Subcellular localization of
targeting signals and transmembrane helices of
deduced protein-coding genes were predicted using
the following programs: PSORT,41 TargetP,42 and
SOSUI.43
2.9. Synteny analysis
Synteny was detected by identifying arrays of pre-
dicted protein-encoding genes between target
genomic regions. Translated amino acid sequences
of the products of genes assigned on the mapped
TGS were compared with those in the reference
genomes, and a BLASTP E-value of ,1e220 was con-
sidered to be significant. Synteny blocks were sur-
veyed on the basis of physically linked sequence
units such as contigs and singlet clones. A synteny
block was defined as the region where three or
more conserved homologs were located within a
100-kb DNA stretch in the two genomes.
3. Results and discussion
3.1. Genome sequencing
Preliminary random and TAC end sequencing
revealed the presence of various types of repetitive
sequences in the L. japonicus genome. The FISH analy-
sis showed that some of these sequences were distrib-
uted along the entire genome with occasional
condensation to different extents, strongly suggesting
that the gene spaces are intermingled with repeated
sequences. In order to efficiently obtain sequence
information for the gene space, we combined two
independent approaches: clone-by-clone sequencing
from seed points of the genome and shotgun sequen-
cing of both STM enriched in gene spaces and a
whole-genomic DNA from which highly repetitive
and organelle genome sequences were subtracted
(SGRs).
A total of 1314 TAC and BAC clones were selected
based on the sequences of ESTs, cDNA and gene infor-
mation from L. japonicus and other legumes, and an
additional 584 TAC and BAC clones were selected by
overlaps. Their nucleotide sequences were determined
according to the shotgun method with three to five
times redundancy (Supplementary Table S1). Further
assembly of the sequences of the 1898 clones
produced 954 supercontigs with a total length
of 167 267 829 bp. Among these sequences, 76 366
532 bp (46%) covered by 823 clones were Phase 3
sequences (HGS). In parallel, 4603 TAC clones,
neither ends of which hit the repetitive sequences,
were pooled and subjected to shotgun sequencing.
Assembly of 808 816 STM random sequences and
847 513 SGR sequences, the sum of which was esti-
mated to give 2.4 times the genome coverage, gener-
ated an additional 109 986 contigs that were 147
805 446 bp long in total (SGA). The efficiency of the
STM and SGR approaches was indicated by the ratio
of centrometic and heterochromatic repeat sequences
in SGA sequences, which was about one-fifth of that in
TAC/BAC end sequences (Supplementary Fig. S1). The
total length of the determined sequences (TGS)
amounted to 315 073 275 bp (Supplementary Table
S2). Although the TGS corresponded to 67% of the
reported L. japonicus genome (472 Mb),44 it can
be estimated that the TGS covers 91.3% of the
gene space because 11 404 out of 12 485 collections
of tentative consensus (TC) sequences of the
L. japonicus Gene Index provided by the Gene Index
Project (http://compbio.dfci.harvard.edu/tgi/plant.
html) were located on the TGS. TGS was used as stan-
dard information for further analysis of gene structure
and function in L. japonicus in this study.
3.2. Construction of the sequence-tagged genetic
linkage map
In order to anchor the obtained genomic sequences
to the genetic linkage map, DNA markers were devel-
oped for the sequenced TAC and BAC clones, and
genetic mapping was carried out. As a result, a total
230 Genome analysis of Lotus japonicus [Vol. 15,

of 788 microsatellite and 80 derived cleaved ampli-
fied polymorphic sequences (dCAPS) markers were
generated and mapped onto the six linkage groups,
which resulted in anchoring of the HGS and the
connecting TAC/BAC and SGR sequences onto the
linkage map. In total, 594 supercontigs containing
130 251 279 bp, corresponding to 41% of the TGS,
were anchored onto the genetic linkage map
(Supplementary Table S1).
Authenticity of the above genetic linkage map was
examined by FISH. TAC clones genetically mapped at
distal and proximal positions to the linkage groups
were used as probes for hybridization. As shown in
Fig. 1, all of the clones examined were successfully
located at the expected positions on the correspond-
ing chromosomes. For chromosome 5, four clones
were positioned inversely in the expected order on
the genetic linkage map; therefore, we revised the
marker order of the genetic map. Genetic and phys-
ical distances did not agree very well, as reported in
A. thaliana and rice.35,36
The positions of centromeres were deduced by
cytological features in both prometaphase and pachy-
tene chromosomes and were further confirmed by
FISH analysis using the pericentromere-specific retro-
element LjRE2 (described later) as a probe (Fig. 1).
TAC clones genetically mapped at distal positions of
each linkage group were located on the telomeric
regions of all the chromosomes by FISH with the
exception of the bottom of chromosome 4 and the
top of chromosome 6 (Fig. 1). However, none of
these clones and the extending sequences contained
the consensus telomeric repeat, CCCTAAA.45 Small
condensed structures were observed at most of the
telomeric regions of the chromosomes.44 The pre-
sence of such heterochromatic regions may have pre-
vented the extension of the analyzed sequences to the
telomeric repeat. Indeed, a short tandem repeat
sequence, LjTR4, specific to the subtelomeric region
was found on the contig (CM0105) at the bottom
of chromosome 1, where no telomeric condentation
was observed.44
3.3. Repetitive sequences
A total of 33 730 di-, tri-, and tetra-nucleotide
simple sequence repeats (SSRs) that were equal to
or longer than 15 bp were identified in the TGS
(Supplementary Table S3). Provided that the size of
the TGS is 315 Mb, the frequency of occurrence of
the above SSRs was estimated to be one SSR in every
9.3 kb. Di-, tri-, and tetra-nucleotide SSRs accounted
for 48.6, 44.4, and 7.0% of the identified SSRs,
respectively. The SSR patterns that appeared
frequently were (AT)n, (AAG)n, and (AAAT)n, each
representing 63% of di-, 28% of tri-, and 46% of
tetra-nucleotide repeat units. The tri-nucleotide
SSRs, particularly (GGT)n and (GGA)n, were preferen-
tially found in exons. (AG)n was enriched in 50 and 30
UTRs, and (AC)n frequently occurred in 50 and 30 UTRs
and introns (Supplementary Table S3).
A search using the repeat sequence finding program
RECON14 against the TGS unraveled the occurrence of
a variety of repeat elements including Class I and Class
Figure 1. FISH detection and the integration map of L. japonicus.
(A and C) Mitotic prometaphase chromosomes and meiotic
pachytene chromosomes were stained with DAPI. (B) TAC
clone, LjT30P03 (corresponding marker: TM0148) (green,
arrowhead) was detected on the long arm of chromosome
5. 45S rDNA (red) was detected on the short arms of
chromosomes 2, 5, and 6. (D) LjT30P03 (corresponding
marker: TM0148) (green, arrowhead) was detected on
chromosome 5 of the pachytene chromosome. The error bar
represents 5 mm. (E) Integration among three maps; mitotic
prometaphase chromosome map, meiotic pachytene
chromosome map, linkage map of L. japonicus. Red circles
show TAC clones, yellow and orange show ribosomal RNA
genes, green shows tandem repeat LjTR1, blue shows
retrotransposon LjRE2 representing the centromere. The length
ratio among six chromosomes was adjusted to the ratio of
pachytene chromosomes.
No. 4] S. Sato et al. 231