DNA adenine methylation changes dramatically during
establishment of symbiosis
Hiroyuki Ichida
1,2
, Tomoki Matsuyama
3
, Tomoko Abe
2
and Takato Koba
1
1 Graduate School of Science and Technology, Chiba University, Matsudo, Japan
2 Accelerator Application Research Group, Nishina Center for Accelerator-Based Science, RIKEN, Hirosawa, Wako, Saitama, Japan
3 Cellular Biochemistry Laboratory, Discovery Research Institute, RIKEN, Hirosawa, Wako, Saitama, Japan
Restriction landmark genome scanning (RLGS) is a
method for the two dimensional display of end-labeled
DNA restriction fragments and is an unbiased method
for DNA methylation scanning in higher eukaryotes
[1]. Virtual image (Vi-) RLGS software simulates two
dimensional DNA electrophoresis patterns based on
whole-genome sequence data, allowing the rapid
matching of DNA spots to their sequences without the
time and effort of cloning [2]. Recent advances in
DNA sequencing technology have enabled the sequen-
cing of entire genomes in various organisms, and the
resulting data allow a comprehensive analysis of gen-
ome dynamics during host–microbe interactions.
Bradyrhizobium japonicum and Mesorhizobium loti
are symbiotic bacteria that perform nitrogen fixation
in host plant roots. Their principal hosts are soybean
and Lotus japonicus, which are an important grain
crop and a model legume, respectively. Biological

The DNA adenine methylation status on specific 5¢-GANTC-3¢ sites and
its change during the establishment of plant–microbe interactions was
demonstrated in several species of a-proteobacteria. Restriction landmark
genome scanning (RLGS), which is a high-resolution two dimensional
DNA electrophoresis method, was used to monitor the genomewide change
in methylation. In the case of Mesorhizobium loti MAFF303099, real
RLGS images obtained with the restriction enzyme MboI, which digests at
GATC sites, almost perfectly matched the virtual RLGS images generated
based on genome sequences. However, only a few spots were observed
when the restriction enzyme HinfI was used, suggesting that most GANTC
(HinfI) sites were tightly methylated and specific sites were unmethylated.
DNA gel blot analysis with the cloned specifically unmethylated regions
(SUMs) showed that some SUMs were methylated differentially in bacter-
oids compared to free-living bacteria. SUMs have also been identified in
other symbiotic and parasitic bacteria. These results suggest that DNA
adenine methylation may contribute to the establishment and ⁄ or mainten-
ance of symbiotic and parasitic relationships.
Abbreviations
CcrM, cell cycle-regulated methyltransferase; Dam, deoxyadenosine methyltransferase; RLGS, restriction landmark genome scanning; SUM,
specifically unmethylated region; Vi, virtual image.
FEBS Journal 274 (2007) 951–962 ª 2007 The Authors Journal compilation ª 2007 FEBS 951
in microbial genomes. In bacteria, these methylated
bases are best known as important agents for restric-
tion-modification systems, which distinguish self and
nonself DNA to protect bacteria from invaders. In this
system, the host DNA is methylated and only unmeth-
ylated DNA is digested by cognate restriction endo-
nucleases [5]. In mammalian genomes, DNA is
methylated at the C5 position of cytosine within CpG
dinucleotide sequences. In the human genome, almost

methylation in diverse bacteria suggests its import-
ance in the regulation of gene expression. However,
the genomewide methylation status has not yet been
elucidated due to the lack of an appropriate analysis
tool.
We used in silico RLGS analysis to achieve genome-
wide monitoring of bacterial DNA methylation
status, successfully demonstrated the existence of
stably unmethlyated regions on several bacterial
genomes, and demonstrated a dramatic change in
methylation during plant–microbe interactions. This
approach may provide novel insights into a variety of
symbiotic and parasitic relationships, including human
diseases.
Results
In silico RLGS visualized the genomes efficiently
We obtained RLGS patterns of A. tumefaciens C58,
B. japonicum NBRC14792, and M. loti MAFF303099
with several enzyme combinations. The most import-
ant step in RLGS analysis is selecting landmark
enzymes that produce well-focused and informative
spot patterns. We used AscI, BspEI, MluI, and NotIas
landmark enzymes. These four enzymes cleave specific
GC-rich sequences and therefore gave satisfactory
resolution of a sufficient number of spots because the
genomes of all three bacterial strains have high GC
content (data not shown). For example, approximately
1071, 979, and 1025 spots were visualized in the RLGS
analysis of M. loti MAFF303099 with AscI, BspEI,
and NotI as landmark enzymes, respectively, in combi-

enzyme combinations without duplication was 87.3%.
This result clearly demonstrates that RLGS analysis
combined with in silico profiling enables efficient and
high density scanning for mutations over the entire
genome with good resolution.
Genomewide analysis of DNA adenine methylation H. Ichida et al.
952 FEBS Journal 274 (2007) 951–962 ª 2007 The Authors Journal compilation ª 2007 FEBS
Adenine methylation status in M. loti
MAFF303099
We scanned for the adenine methylation status of
M. loti MAFF303099 using in silico RLGS profiling.
The a-proteobacteria, including M. loti MAFF303099,
are thought to have CcrM, which transfers a methyl
group from S-adenosylmethionine to the amino group
of the adenine moiety embedded in the sequence
5¢-GANTC-3¢. The restriction endonuclease HinfI
cleaves unmethylated GANTC sites, but not methyl-
ated sites. The cleavage by the landmark enzymes
(AscI, MluI, and NotI) is not affected by CcrM
methylation because of the lack of GANTC sequences
on their recognition sequences; therefore, the spot
intensity directly reflected the CcrM methylation status
at the HinfI site of the corresponding genome region.
The deduced total coverage with these three enzyme
combinations was 91.0% (Table 1).
The real and virtual RLGS patterns of M. loti
MAFF303099 obtained with NotI–HinfI are shown in
Fig. 3. Similar results were obtained with AscI–HinfI
and MluI–HinfI (data not shown). Most of spots on
the real RLGS patterns clustered on the top (Fig. 3A;

The large divisions on the scale indicate
1 Mb.
Table 1. RLGS coverage (%) in the three bacteria. 1D, Percentage of genome regions that visualize in first dimensional (agarose gel) electro-
phoresis. 2D, Percentage of genome regions that visualize in second dimensional (polyacrylamide gel) electrophoresis.
Enzyme combination
and dimension
Agrobacterium
tumefaciens C58
Circular chr.
Agrobacterium
tumefaciens C58
Linear chr.
Bradyrhizobium
japonicum
USDA110
Mesorhizobium
loti MAFF303099
1D 2D 1D 2D 1D 2D 1D 2D
AscI-Mbol 27.4 1.3 23.8 1.0 39.3 2.0 42.4 2.4
BspEl-Mbol 70.0 5.8 68.2 5.8 56.4 4.4 61.8 4.5
Notl-Mbol 34.2 2.0 41.8 2.2 48.8 3.2 49.2 3.2
Total coverage 85.9 8.7 86.7 8.6 86.1 8.9 87.3 9.5
Ascl-Hinfl 26.9 2.7 21.6 2.0 58.5 7.8 52.6 7.2
Mlul-Hinfl 51.3 6.5 48.9 6.2 68.5 9.2 55.9 7.0
Notl-Hinfl 36.5 4.1 42.6 4.5 70.8 12.1 61.0 8.9
Total coverage 77.0 12.5 80.0 12.3 95.6 20.1 91.0 21.0
Fig. 3. Real and virtual RLGS patterns obtained using NotI–HinfI. (A) Real RLGS pattern of M. loti MAFF303099. The image was obtained
using NotI as the landmark enzyme and HinfI as the second dimension fragmentation enzyme. Although most of the spots are located on
the top left (short in the first dimension and long in the second dimension), 104 apparent spots are visualized below the cluster. (B) Virtual
RLGS pattern calculated based on the whole-genome sequence and conditions corresponding to those in (A). Unlike in the real RLGS pat-

and cleavage by HinfI was blocked due to methylation
on adenine and ⁄ or cytosine residues on the GANTC
sites. Bisulfite sequencing, which is a widely used tech-
nique to determine cytosine methylation levels in base
pair resolution, demonstrated that the cytosine nucleo-
tides, including HinfI sites, were completely unmethyl-
ated (data not shown). Therefore, the blocking of
HinfI cleavage was caused by DNA adenine methyla-
tion at GANTC sites. Interestingly, 82, 94, and 104
spots were observed in the correct second dimension
position in the images obtained with AscI–HinfI,
MluI–HinfI, and NotI–HinfI, respectively (Fig. 3A,B,
and data not shown). These spots suggest that the
M. loti MAFF303099 genome was partly unmethyl-
ated, and the methylation status had been inherited
stably. We refer to these regions as specifically un-
methylated regions (SUMs).
Comprehensive catalog of specifically
unmethylated regions
Real and virtual RLGS images obtained with HinfI
clearly demonstrated the occurrence of SUMs; how-
ever, their nucleotide sequences could not be obtained
by in silico RLGS profiling because most of the spots
were methylated and the number of informative land-
mark spots was insufficient for matching the real and
virtual RLGS patterns. There were 104 spots on the
real NotI–HinfI RLGS pattern and its coverage was
61.0% (Fig. 3A and Table 1). Therefore, we estimated
that there were  170 nonredundant SUMs in the
free-living M. loti MAFF303099 genome.

ules consist of a mixture of plant and bacterial cells;
therefore, uninfected plant root DNA was used as a
negative control for the nodule pattern (data not
shown). Comparisons between free-living M. loti
MAFF303099 and bacteroids with
AscI–HinfI, MluI–
HinfI, and NotI–HinfI revealed that the signal intensi-
ties derived from SUMs were decreased distinctly in
bacteroids (Fig. 3E,F, and data not shown; arrow-
heads indicate some examples of spots that disap-
peared); these patterns were obtained under the same
conditions, i.e., the amount of DNA, reagent lots, elec-
trophoresis, autoradiography, and film development
and digitization. These results suggest that the bacter-
ial DNA adenine methylation status changes during
the establishment of the symbiotic relationship and
may contribute to the regulation of plant–microbe
interactions.
To confirm the change in methylation during nodule
development, DNA from free-living bacteria and nod-
ules was probed with the cloned SUMs described
above (Table 2). Although we collected 145 individual
SUMs, fragments less than 200 bp in length did not
provide sufficient sensitivity. Therefore, the 29 SUMs
that were longer than 200 bp in length and appeared
twice or more were chosen as probes. Of these, 27
(93.1%) gave significant signals in the NotI–HinfI
digestion, but not in the NotI digestion. The signal
intensity of 20 loci was decreased in nodules; therefore,
these loci are methylated during nodule development

combination. These results suggest that SUMs are
widely distributed in a variety of bacteria and may
play a significant role in regulatory mechanisms.
Discussion
We showed that in silico RLGS profiling, which is
based on a comparison between real RLGS patterns
Table 2. DNA methylation levels at the SUMs in free-living and bacteroids, determined by DNA gel blot analysis. The unmethylated HinfI is
underlined. Unmethylated band intensity was expressed as – (unmethylated signal not detected) and + to +++++(weakest to strongest
signal). Asterisks indicate two or more unmethylated bands were detected.
Identifier
Length
(bp) GC (%)
Corresponding genome region Unmethylated band intensity
Replicon Position HinfI flanking sequence Free-living Nodule
ARM-AH1-A01 747 52.0 Chr 4 832 908–4 833 663
a
CCATTTCA GAGTC GATGGGAC ++
ARM-AH1-A05 310 63.6 Chr 6 203 593–6 203 911 TTTCGCGG
GATTC TATGGTGA +++ –
ARM-AH1-B07
b
600 54.5 Chr 5 243 365–5 243 973
a
GCCAAGAA GATTC TGTGGTCG ++++ +++
ARM-AH1-B09
b
315 53.1 Chr 5 028 568–5 028 891
a
CAAATTGC GACTC AGGACGTT + –
ARM-AH1-C07

a
AGAGAATA GAGTC GTGTATTA +++ –
ARM-NH1-F05 246 54.9 Chr 3 720 571–3 720 825 AGGACCAT
GATTC GGACTTGG ++,* ++,*
ARM-NH1-H12 288 51.2 Chr 4 916 344–4 916 637
a
ATAATGAA GAGTC GTTCATCC ++,* +
ARM-NH2-B02 393 53.0 Chr 5 334 423–5 334 780 TCAGTCAT
GAGTC ACTCCGAA ++,* ++,*
ARM-NH2-D07 586 57.1 Chr 5 443 356–5 443 950 CGTCGAGC
GATTC CCAAGTTT ++++ ++
ARM-MH1-D03 246 55.3 pMLa 153 845–154 098 GTCCCCTA
GATTC CACTTTAT +++ +
ARM-MH1-D05 297 54.6 pMLa 56 482–56 786 TTCCCGGA
GAATC GTCAAATT ++ ++
ARM-NH1-A10 528 55.3 pMLa 105 514–106 050 TTTATATC
GAGTC TGTTACGG +++++ +++
ARM-NH1-E12
b
555 57.6 pMLa 310 685–311 238 TACGTTTT GAGTC TGCAACAT +++++ +
ARM-NH1-F08 583 58.1 pMLa 154 241–154 803 GCTTGTAG
GATTC ACTTCAAA +++++,* +,*
ARM-AH1-A06 434 57.6 pMLb 113 426–113 868 AATTCCCT
GACTC CCGTCGAA + + + + +, * + +, *
ARM-NH1-B05
b
566 56.2 pMLb 184 911–185 482 GTCTGTTC GAGTC AGCAAACC ++++ ++,*
ARM-NH1-F06 261 58.1 pMLb 46 255–46 524 TTGATTAA
GATTC CTAATTTA ++++,* +
a

CcrM target sequences, are usually methylated in the
M. loti MAFF303099, B. japonicum NBRC14792, and
A. tumefaciens C58 genomes (Fig. 3). However, some
GANTC sites in these genomes are specifically
unmethylated, and the methylation status is heritable.
We obtained 145 nonredundant SUMs from 339 indi-
vidual clones using adapter-mediated PCR (Table S2),
which may comprise  85% of all SUMs in the gen-
ome of free-living M. loti MAFF303099. Sequencing
and mapping results suggest that the SUMs are loca-
lized in low-GC regions on the genome, and their aver-
age GC content was much lower than those of the
main chromosome and two plasmids (Fig. 2 and
Table 2). Horizontal gene transfer is thought to be a
major force in genome plasticity and may play a cru-
cial role in evolution [16]. Historically, fitness-enhan-
cing traits such as antibiotic resistance, virulence,
organic solvent degradability, and symbiotic nitrogen
fixation ability were transmitted by this mechanism.
Now, many gene candidates transferred among pro-
karyotes and from prokaryotes to eukaryotes are iden-
tified via comparative genomic analysis [17]. The
relationship between SUMs and the GC content indi-
cates an association with genome evolution.
Why and how are some genome regions specifically
unmethylated? The pioneering work conducted with
Dam of Escherichia coli hints at the answers to these
questions. Dam belongs to the a group of methyl-
transferases and transfers a methyl group to the N6
position of the adenine in 5¢-GATC-3¢ sites. Tavazoie

mll7872 (position 6 514 807–6 513 713; encodes an
unknown protein) and mlr7873 (position 6 515 256–
6 517 130; encodes a cellulose synthase-like protein),
respectively. The other motif was a repeat core unit of
nod box (ATC-N
9
-GAT [20]), which is the binding site
of NodD, a LysR-type transcriptional regulator that
directs specific flavonoid-dependent nodulation gene
expression [21]. Of the 145 nonredundant SUM
sequences, 16 contained this core motif and seven of
these were located on the symbiosis island (Table 2
and S2). The number of nucleotides between the nod
repeat core unit and the unmethylated Hin fI site varied
from 1 to 681, and averaged 283. Although most
NodD proteins bind to the promoter when specific
flavonoids are present, some are activated independ-
ently of flavonoids and have greater transcriptional
activity than the flavonoid-dependent proteins [22]. In
addition, the flavonoid-dependent NodD proteins also
exhibit relatively weaker, but detectable, DNA-binding
activity in the absence of inducers [23]. Therefore,
at least some type of NodD protein can competi-
tively inhibit CcrM methylation, even in the free-living
condition. It is likely that SUMs are formed by
competition between CcrM and DNA-binding
proteins. Biochemical analysis with purified CcrM and
various DNA-binding proteins will be a key for further
analysis.
This is the first report of specific unmethylation of

chased from ATCC via an official local distributor (Summit
Pharmaceuticals International, Tokyo, Japan).
Lotus japonicus MG-20 seeds were a gift from the
National Bio Resource Project (Miyazaki University,
Miyazaki, Japan). Surface-sterilized L. japonicus MG-20
seeds were germinated on B & D nitrogen-free plates [25]
under a photoperiod of 16 h light ⁄ 8 h dark at 22 °C. A
log-phase culture of M. loti MAFF303099 (optical density
at 600 nm, 0.4–0.6) was washed three times with sterilized
distilled water. L. japonicus MG-20 seedlings with roots 15–
20 mm long were soaked in the washed bacterial cell sus-
pension for 1 min. The inoculated plants were placed on
new B & D nitrogen-free plates and grown for 45 days.
Using this method, one to three nodules usually developed
on each plant. The nodules on green plants with elongated
shoots were harvested as nitrogen-fixing nodules and used
for the following experiments.
DNA protocols
General molecular manipulations were carried out accord-
ing to standard procedures, unless otherwise specified. Bac-
terial DNA was extracted from fresh log-phase cultures
(optical density at 600 nm, 0.4–0.6) using the cetyltrimethyl-
ammonium bromide procedure [26]. Nodule DNA and
plant root DNA were extracted from 100 mg of tissue using
a Nucleon PhytoPure DNA extraction kit according to the
manufacturer’s instructions (GE Healthcare Bio-Sciences,
Piscataway, NJ, USA). All oligonucleotide sequences used
are listed in Table S1.
H. Ichida et al. Genomewide analysis of DNA adenine methylation
FEBS Journal 274 (2007) 951–962 ª 2007 The Authors Journal compilation ª 2007 FEBS 959

B. japonicum USDA110, and M. loti MAFF303099 repor-
ted previously [28–30]. ‘Coverage’ was defined as the
percentage of the total length of visualized spots in the
RLGS images relative to the genome size. For example,
‘first dimension coverage’ of an RLGS image of M. loti
MAFF303099 was calculated as (sum of first dimension
length on the image ⁄ size of the main chromosome:
7 036 071) · 100. The first and second dimension coverages
were calculated computationally from the in silico simula-
tion of RLGS reactions using the genome sequences. The
spots located between 500 and 15 000 bp in the first dimen-
sion and 100 and 1000 bp in the second dimension were
counted as ‘visualized’ spots. Genome regions visualized
with two or more combinations of enzymes were counted
only once when calculating the total coverage from plural
RLGS images.
Methylation profiling
SUMs of the M. loti MAFF303099 genome were amplified
using adapter-mediated PCR. The adapter was synthesized
as two individual oligonucleotides and annealed by boiling
for 3 min, followed by gradual cooling to room tempera-
ture (Table S1). One microgram of M. loti MAFF303099
DNA was digested with 10 units (U) of a landmark enzyme
(NotI, AscI, or MluI) and HinfI for 3 h and recovered by
ethanol precipitation. The precipitated DNA was dissolved
in 3 lL of water and added into the ligation mixture
[100 ng each of landmark- and HinfI-adapter and 5 lLof
Ligation High solution (Toyobo, Tokyo, Japan) in 8.5 lL].
The mixture was incubated at 16 °C for 16 h and subjected
to PCR amplification without purification. PCR was per-

+
mem-
branes (GE Healthcare Bio-Sciences). Probes were prepared
by PCR amplification of the insert regions of each clone
using the primers CasA-specific and CasB-specific. Labeling
and detection were performed using an ECL direct nucleic
acid labeling and detection system (GE Healthcare Bio-
Sciences) according to the manufacturer’s instructions.
Hybridization was performed overnight in the supplied
hybridization buffer containing 0.1 m sodium chloride at
42 °C.
Quantification of plasmid copy numbers
Copy numbers of the two plasmids, pMLa and pMLb, of
M. loti MAFF303099 under free-living and bacteroid con-
ditions were determined by real-time PCR. Three primer
pairs, which amplify evenly distributed regions on the tar-
get, were designed for each replicon. The reaction mixture
consisted of 25 lL of SYBR premix Ex Taq (Takara),
10 pmol of primers, and 10 lL of diluted template DNA in
50 lL. Amplification and real-time quantification were per-
formed with 40 cycles of 94 °C for 30 s, 60 °C for 30 s, and
Genomewide analysis of DNA adenine methylation H. Ichida et al.
960 FEBS Journal 274 (2007) 951–962 ª 2007 The Authors Journal compilation ª 2007 FEBS
72 °C for 20 s using an ABI 7900HT (Applied Biosystems).
Specific amplification of the targets was confirmed by
melting curve analysis and 10% (w ⁄ v) polyacrylamide gel
electrophoresis.
Acknowledgements
The authors thank Sumie Ohbu (Nishina Center for
Accelerator-Based Science, RIKEN) for her technical

memory. Genes Dev 16, 6–21.
7 Jones PA & Baylin SB (2002) The fundamental role of
epigenetic events in cancer. Nat Rev Genet 3, 415–428.
8 Riggs AD & Pfeifer GP (1992) X-chromosome inactiva-
tion and cell memory. Trends Genet 8, 169–174.
9 Swain JL, Stewart TA & Leder P (1987) Parental legacy
determines methylation and expression of an autosomal
transgene: a molecular mechanism for parental imprint-
ing. Cell 50, 719–727.
10 Zweiger G, Marczynski G & Shapiro L (1994) A Caulo-
bacter DNA methyltransferase that functions only in
the predivisional cell. J Mol Biol 235, 472–485.
11 Ratel D, Ravanat J, Berger F & Wion D (2006)
N6-methyladenine: the other methylated base of DNA.
Bio Essays 28, 309–315.
12 Low DA, Weyend NJ & Mahan MJ (2001) Roles of
DNA adenine methylation in regulating bacterial gene
expression and virulence. Infect Immun 69, 7197–7204.
13 Stephens CM, Zweiger G & Shapiro L (1995) Coordi-
nate cell cycle control of a Caulobacter DNA methyl-
transferase and the flagellar genetic hierarchy.
J Bacteriol 177, 1662–1669.
14 Robertson GT, Reisenauer A, Wright R, Jensen RB,
Jensen A, Shapiro L & Roop RMI (2000) The Brucella
abortus CcrM DNA methyltransferase is essential for
viability, and its overexpression attenuates intracellular
replication in murine macrophages. J Bacteriol 182,
3482–3489.
15 Reisenauer A, Kahng LS, McCollum S & Shapiro L
(1999) Bacterial DNA methylation: a cell cycle regula-

tion of nodulating peanut rhizobia isolated from native
soil population in Cordoba, Argentina. Symbiosis 33,
59–72.
25 Broughton WJ & Dilworth MJ (1971) Control of leghe-
moglobin synthesis in snake beans. Biochem J 125,
1075–1080.
26 Wilson K (1994) Preparation of genomic DNA from
bacteria. In Current Protocols in Molecular Biology, 4th
edn. (Ausubel FM, Brent R, Kingston RE, Moore DD,
H. Ichida et al. Genomewide analysis of DNA adenine methylation
FEBS Journal 274 (2007) 951–962 ª 2007 The Authors Journal compilation ª 2007 FEBS 961
Seidman JG, Smith J A & Strauhl K, eds), pp. 2.4.1–
2.4.2, John Willy & Sons Inc, Hoboken, NJ.
27 Hatada I, Hayashizaki Y, Hirotsune S, Komatsubara H
& Mukai T (1991) A genomic scanning method for
higher organisms using restriction sites as landmarks.
Proc Natl Acad Sci USA 8, 9523–9527.
28 Goodner B, Hinkle G, Gattung S, Miller N, Blanchard
M, Qurollo B, Goldman BS, Cao Y, Askenazi M,
Halling C et al. (2001) Genome sequence of the plant
pathogen and biotechnology agent Agrobacterium tume-
faciens C58. Science 294, 2323–2328.
29 Kaneko T, Nakamura Y, Sato S, Asamizu E, Kato T,
Sasamoto S, Watanabe A, Idesawa K, Ishikawa A,
Kawashima K et al. (2000) Complete genome structure
of the nitrogen-fixing symbiotic bacterium Mesorhizo-
bium loti. DNA Res 7, 331–338.
30 Kaneko T, Nakamura Y, Sato S, Minamisawa K,
Uchiumi T, Sasamoto S, Watanabe A, Idesawa K,
Iriguchi M, Kawashima K et al. (2002) Complete geno-