Deficiency in the amino aldehyde dehydro

An assay of AMADH activity in soybeans revealed that the aromatic soybean, which contains 2AP, also lacked AMADH enzyme activity.. We chose to study the above hypotheses in the soybean f

Deficiency in the amino aldehyde dehydrogenase

encoded by GmAMADH2, the homologue of rice Os2AP, enhances 2-acetyl-1-pyrroline biosynthesis in soybeans (Glycine max L.)

Siwaret Arikit1,2,†, Tadashi Yoshihashi3,†, Samart Wanchana4, Tran T Uyen3, Nguyen T T Huong3, Sugunya Wongpornchai5and Apichart Vanavichit1,6,*

1 Rice Science Center and Rice Gene Discovery, Kasetsart University Kamphaeng Saen Campus, Nakhon Pathom, Thailand

2 Interdisciplinary Graduate Program in Genetic Engineering, Kasetsart University, Bangkok, Thailand

3 Postharvest Science and Technology Division, Japan International Research Center for Agricultural Sciences, Tsukuba Ibaraki, Japan

4 International Rice Research Institute, Los Ban˜os, Laguna, Philippines

5 Department of Chemistry, Faculty of Science, Chiang Mai University, Chiang Mai, Thailand

6 Department of Agronomy, Kasetsart Univerisity Kamphaeng Saen, Nakhon Pathom, Thailand

Received 30 September 2009;

accepted 5 April 2010.

*Correspondence (fax +66 34 355197;

e-mail vanavichit@gmail.com)

† These authors contributed equally to this

work.

Keywords: AMADH,

2-acetyl-1-pyrr-oline, vegetable soybean, GABA,

polyamine metabolism, Os2AP.

Summary 2-Acetyl-1-pyrroline (2AP), the volatile compound that provides the ‘popcorn-like’ aroma in a large variety of cereal and food products, is widely found in nature Defi-ciency in amino aldehyde dehydrogenase (AMADH) was previously shown to be the likely cause of 2AP biosynthesis in rice (Oryza sativa L.) In this study, the validity of this mechanism was investigated in soybeans (Glycine max L.) An assay of AMADH activity in soybeans revealed that the aromatic soybean, which contains 2AP, also lacked AMADH enzyme activity Two genes, GmAMADH1 and GmAMADH2, which are homologous to the rice Os2AP gene that encodes AMADH, were characterized The transcription level of GmAMADH2 was lower in aromatic varieties than in nona-romatic varieties, whereas the expression of GmAMADH1 did not differ A double nucleotide (TT) deletion was found in exon 10 of GmAMADH2 in all aromatic varie-ties This variation caused a frame-shift mutation and a premature stop codon Sup-pression of GmAMADH2 by introduction of a GmAMADH2-RNAi construct into the calli of the two nonaromatic wild-type varieties inhibited the synthesis of AMADH and induced the biosynthesis of 2AP These results suggest that deficiency in the GmAMADH2 product, AMADH, plays a similar role in soybean as in rice, which is to promote 2AP biosynthesis This phenomenon might be a conserved mechanism among plant species

Introduction

2-Acetyl-1-pyrroline (2AP) is a volatile compound that

produces the potent ‘popcorn-like’ or ‘pandan-like’ aroma

found in a large variety of cereal products and

vegetable-derived and animal-vegetable-derived products (Adams and De

Kimpe, 2006) This aroma component is a value-added

characteristic and can be directly linked to consumer

preference in positive terms (Fitzgerald et al., 2009) In rice, the aroma is considered to be a special trait that enhances aromatic rice such as the Jasmine types of Thai-land and Basmati rice of India and Pakistan This aroma is economically important because it determines the pre-mium price Similarly, in vegetable soybean, the immature seeds are consumed as a vegetable or snack called ‘Edam-ame’, and the aroma plays an important role influencing

consumer preference and acceptance ‘Chamame’, a

spe-cial group of ‘Edamame’ that contains a pleasant aroma

that is thought to be attributable to the presence of 2AP

(Fushimi and Masuda, 2001), is preferred by consumers

and can command higher prices In Japan, the price of this

variety can be double that of nonaromatic varieties

(Statis-tics Department, Ministry of Agriculture, Forestry and

Fish-eries, 2009) The aroma characteristic also contributes to

the acceptance of the product by the unfamiliar

con-sumer Hence, the aromatic varieties were targeted as the

first priority in vegetable soybean breeding programmes

that were introduced to 107 countries all over the world

(Shanmugasundaram and Yan, 2004; Takahashi et al.,

2006)

2AP has been reported to be biosynthesized in various

organisms, including plants, such as pandan (Pandanus

amaryllifolius Roxb.) (Buttery et al., 1983), certain varieties

of rice (Oryza sativa L.) (Buttery et al., 1982; Widjaja et al.,

1996; Yoshihashi, 2002), bread flowers (Vallaris glabra

Ktze) (Wongpornchai et al., 2003) and certain varieties of

soybean (Glycine max L.) (Fushimi and Masuda, 2001;

Plonjarean et al., 2007; Wu et al., 2009) Synthesis of this

compound has also been documented in microorganisms,

in particular Bacillus cereus (Adams and De Kimpe, 2007)

and bakers’ yeast (Snowdon et al., 2006), and in animals

such as tigers (Panthera tigris tigris) (Brahmachary et al.,

1990) 2AP can also be formed in food products, such as

roasted popcorn and bread crust, by thermal generation

during heating (Schieberle and Wener, 1991)

So far, the complete biosynthetic pathway for 2AP in

plants has not been fully elucidated Recently, a key

path-way for 2AP biosynthesis in rice was proposed (Bradbury

et al., 2008), and a single gene responsible for the aroma

trait has been identified by different research groups

(Bradbury et al., 2005; Vanavichit et al., 2008; Kovach

et al., 2009) The identified gene has been given multiple

names, including Os2AP (Vanavichit et al., 2008), BAD2

(Bradbury et al., 2005, 2008) and BADH2 (Niu et al.,

2008) Its protein product was identified as amino

alde-hyde dehydrogenase (AMADH), which shares high

sequence similarity with betaine aldehyde dehydrogenase

(BADH) (Bradbury et al., 2008) The utilization of

4-amino-butanal by the AMADH enzyme is likely to be the key step

that affects the biosynthesis of 2AP in rice Nonaromatic

rice contains the functional AMADH enzyme, which

con-verts 4-aminobutanal to 4-aminobutyrate (GABA), and

consequently, 2AP is not formed In aromatic rice, which

does not have a functional AMADH enzyme,

4-aminobut-anal is converted first into precursor(s) and then into the

end product, 2AP, by an unknown pathway (Bradbury

et al., 2008)

Plant AMADHs belong to the aldehyde dehydrogenase (ALDH) superfamily, which represents a group of NAD(P)+ -dependent enzymes that oxidize various aldehydes (Peroz-ich et al., 1999) Based on the amino acid sequences, AMADHs are homologous to BADHs and contain similar primary structures (Brauner et al., 2003) However, there are some differences between the two enzymes with respect to substrate specificities (Sˇebela et al., 2000; Livingstone et al., 2002) AMADH probably plays a role in physiological processes connected to polyamine degrada-tion, converting 4-aminobutanal to GABA (Petrˇivalsky´

et al., 2007) Because natural substrates of AMADH are reactive metabolites that show considerable toxicity, this enzyme was thought to serve as a detoxification enzyme (Tylichova´ et al., 2007) The pathway of GABA biosynthesis via polyamine catabolism, 4-aminobutanal and AMADH is normally thought of as an alternative (Kakkar et al., 2000), while the major pathway occurs via the decarboxylation of glutamate (Snedden et al., 1995) However, evidence sup-porting the existence of this alternative pathway in plants has previously been reported in soybean (Xing et al., 2007) and pea (Petrˇivalsky´ et al., 2007)

At present, however, it has only been reported in rice that AMADH deficiency likely results in 2AP biosynthesis There is no evidence so far as to whether the same mech-anism or biochemical pathway for 2AP biosynthesis is shared among other plants In this study, we aimed to ver-ify the association between AMADH deficiency and 2AP biosynthesis in the soybean to examine its role in a plant besides rice We predicted that inactivation of AMADH, which induces 2AP biosynthesis, would be conserved among plant species and that the gene responsible for AMADH could be orthologous or homologous to the rice Os2AP We chose to study the above hypotheses in the soybean for the following reasons: first, aromatic and nonaromatic varieties of this plant are available, so the natural variation in the gene that controls this trait could

be observed; second, the complete genome sequence of the soybean has been released, allowing us to search for homologous genes throughout its genome; and finally, soybean is a dicot species distantly related to rice, which strengthens the validity of our hypotheses We character-ized the AMADH candidate gene in soybean and inacti-vated it by RNAi to disrupt the synthesis of AMADH and observe its effect on 2AP biosynthesis The results pre-sented here support the idea that inactivation of AMADH

is a general key factor that determines 2AP biosynthesis in

plants These results could be applied to the manipulation

of the target gene in other plants to enhance 2AP

biosynthesis

Results

Determination of the aroma component and

assessment of AMADH activity in aromatic and

nonaromatic soybeans

Ten varieties of aromatic and nonaromatic soybeans were

selected to assay AMADH enzymatic activity in immature

seeds The aroma characteristic of the ten varieties was

confirmed by measuring the 2AP content in the seeds

using headspace gas chromatography (HSGC) 2AP was

detected in all five aromatic varieties but not in the

nona-romatic varieties (Table 1) This result shows that 2AP is

the determinant of the aromatic phenotype in soybeans

The AMADH enzymatic activity of the fractionated extracts

from immature seeds was assayed by activity staining after

native polyacrylamide gel electrophoresis (PAGE)

4-Aminobutanal and betaine aldehyde were compared as

substrates for AMADH enzymatic activity The extracts

from nonaromatic rice (c.v Nipponbare) calli were also

loaded in the same PAGE gel The activity staining test

revealed that AMADH from soybean extracts utilized

4-aminobutanal but could not utilize betaine aldehyde as

a substrate On the contrary, AMADH from rice extracts,

which was used as a positive control, utilized both

4-aminobutanal and betaine aldehyde (Supplementary

Figure S1) Thus, 4-aminobutanal was used as a substrate

for the analysis of AMADH enzymatic in extracts from the

ten soybean varieties The enzyme activity staining on the PAGE gel revealed four activity bands for AMADH with 4-aminobutanal, but the bands could not utilize betaine aldehyde as a substrate (Figure 1) Among these four activity bands, one band (band c, Figure 1) was present in all varieties of nonaromatic soybean but absent from all aromatic soybean varieties This band is considered to be the band of AMADH associated with 2AP production

Characterization of rice Os2AP homologues in soybeans

We hypothesized that the gene encoding the AMADH that is associated with 2AP in soybeans could be the homologue of rice Os2AP We then identified the possible homologues in soybeans using the protein sequence of Os2AP (Vanavichit et al., 2008) to perform a homology search against the NCBI protein database using the Basic Local Alignment Search Tool for protein searching (BLASTP) Two soybean proteins, accession numbers BAG09376 and BAG09377, were retrieved The amino acid sequence identities of BAG09376 and BAG09377 in comparison with rice Os2AP were 75% and 74%, respec-tively Both BAG09376 and BAG09377 were annotated as peroxisomal BADH proteins that were identified in peroxi-somes purified from etiolated soybean cotyledons (Arai et al., 2008)

To obtain the full-length genomic sequences of the genes, the two coding sequences (CDSs), AB333793 and AB333794, corresponding to BAG09376 and BAG09377 were used to perform a nucleotide search (BLASTN) against a recently released (assembly Glyma1) shotgun genome sequence database for soybean (Phytozome, http://www.phytozome.net/soybean) Two gene models, Glyma05g01770 and Glyma06g19820, annotated as BADHs, were retrieved corresponding to AB333793 and

Table 1 Contents of 2AP in soybean seeds determined by

auto-mated headspace gas chromatography with a

nitrogen–phospho-rus detector (HSGC-NPD)

Variety

2AP content (ppb) (mean ± SD) Phenotype

Kaori hime 1160.0 ± 50.4 Aromatic

Fukunari 609.4 ± 43.1 Aromatic

Yuagari musume 1008.5 ± 56.8 Aromatic

n.d., not detected.

Figure 1 AMADH activity assay on a native PAGE gel of crude extracts from soybean seeds from ten varieties The arrows indicate the four bands of enzymes that metabolize 4-aminobutanal Nonaro-matic varieties are Okuhara wase in lane 1, Oishi Edamame in lane 2, Shirono Mai in lane 3, Chiang Mai 60 in lane 4 and Jack in lane 5 Aromatic varieties are Chamame in lane 6, Kouri in lane 7, Kaori hime

in lane 8, Fukunari in lane 9 and Yuagari musume in lane 10.

AB333794, respectively We renamed the two genes

GmAMADH1 and GmAMADH2 in this study and

consid-ered both of them to be candidate genes for AMADH

Both GmAMADH1 and GmAMADH2 contain 15 exons,

similar to other previously reported plant BADHs (Legaria

et al., 1998; Vanavichit et al., 2008) The exons in the two

genes were relatively similar in size, but several introns of

GmAMADH1 were longer than those of GmAMADH2

(Figure 2a) The CDSs and deduced amino acid sequences

of GmAMADH1 and GmAMADH2 from the Phytozome soybean genome database were identical to those reported previously (Arai et al., 2008) GmAMADH1 and GmAMADH2 are 85% identical and 92% similar at the amino acid level and 85% identical at the nucleotide level The essential aldehyde dehydrogenase conserved domain and glutamic acid (Glu) and cysteine (Cys) active site resi-dues were found in both GmAMADH1 and GmAMADH2 (data not shown)

(a)

(b)

Figure 2 (a) The gene structures of GmA-MADH1 and GmAMADH2 The sequence var-iation, a TT deletion in exon 10 of

GmAMADH2 at position 928, is shown (b) Deduced amino acid sequence of the GmA-MADH2 coding sequence translated from the start codon (ATG) The end of translation and the location of the premature stop codon (TAG) are indicated with asterisks (*).

Expression analysis of GmAMADH1 and GmAMADH2

The expression of the two genes, GmAMADH1 and

GmA-MADH2, was analysed in leaves, young seeds and calli,

and these results were compared among the ten aromatic

and nonaromatic soybean varieties The results of reverse

transcription polymerase chain reaction (RT-PCR) analysis

indicated that the expression patterns of the two genes

differed among these soybean varieties The expression

levels of GmAMADH2 were much lower in all tissues of

the five aromatic varieties than in the nonaromatic

varie-ties By contrast, the expression levels of GmAMADH1 did

not differ in all ten soybean varieties (Figure 3) This result

suggested that the lower expression of GmAMADH2 in

the aromatic varieties was associated with the presence of

2AP

Sequence variation in the CDS of GmAMADH2

Because the expression of GmAMADH2 was lower in all

of the aromatic soybean varieties compared with the

nonaromatic varieties, the factors that regulate the

expres-sion of this gene were investigated The full-length

GmA-MADH2 cDNA was amplified by RT-PCR from two

representative soybean varieties, Chamame (aromatic) and

CM60 (nonaromatic), and it was subsequently sequenced

Pairwise alignment of the cDNA sequences between the

two varieties revealed that the CDSs were almost identical,

except for a region in exon 10 at nucleotides 928–932

downstream from the first ATG, where two thymines (T)

were absent from the aromatic variety Chamame

(Figure 2a) The TT deletion in Chamame caused a

frame-shift and a premature stop codon, TAG, three bases

downstream from the deletion (Figure 2b) Consequently,

the deduced amino acid sequence from the CDS of

Cham-ame was truncated and comprised only 311 amino acids

By contrast, the nonaromatic variety, CM60, contained

the complete 503 amino acid sequence, and this sequence

was identical to that of BAG09376 The presence of the 2-bp deletion in exon 10 in the other four aromatic varie-ties was confirmed by sequencing exon 10 of the genomic DNA (Figure 2a) This result indicated that the lower amount of the GmAMADH2 transcript observed in all aro-matic varieties could be caused by the premature stop codon that induced non-sense-mediated decay (NMD) (Is-shiki et al., 2001)

RNAi-mediated gene suppression of GmAMADH2

To verify that the AMADH associated with 2AP biosynthe-sis (band c, Figure 1) was encoded by GmAMADH2 and that the level of 2AP would increase when the AMADH was inactivated, we suppressed the expression of GmA-MADH2 in nonaromatic soybeans by RNAi-mediated gene suppression The GmAMADH2-RNAi was introduced into two nonaromatic soybean varieties, CM60 and Jack, using Agrobacterium transformation The pANGmAMADH2 con-struct contained genomic DNA fragments of GmAMADH2

in the sense and antisense directions that spanned 441 bp covering exon 1, intron 1 and part of exon 2 (Figure 4a)

To select for the GmAMADH2-RNAi-containing calli, all hygromycin-resistant calli were tested by PCR to detect the GUS linker fragment that served as the loop in the transcribed RNAi construct In this study, we investigated the level of gene suppression, 2AP content and enzyme activity at the callus stage The expression levels of GmA-MADH2 in pANGmAGmA-MADH2-transformed calli of both varieties were highly suppressed compared with the levels

of the wild type (Figure 4b) The expression of GmA-MADH1 was also analysed to verify the specificity of the RNAi vector The results showed that the expression of GmAMADH1 was not suppressed in the two RNAi-trans-formed lines, indicating that the pANGmAMADH2 RNAi vector was specific for GmAMADH2 without co-suppres-sion of GmAMADH1 The accumulation of the siRNA gen-erated by the RNAi mechanism was observed by RNA gel

Figure 3 RT-PCR analysis of GmAMADH1

and GmAMADH2 expression in leaves, young

seeds and calli Lectin was used as a control.

Nonaromatic varieties are Okuhara wase in

lane 1, Oishi Edamame in lane 2, Shirono

Mai in lane 3, Chaing Mai 60 in lane 4 and

Jack in lane 5 Aromatic varieties are

Chamame in lane 6, Kouri in lane 7, Kaori

hime in lane 8, Fukunari in lane 9 and

Yuagari musume in lane 10.

blot analysis of pANGmAMADH2-transformed calli of both

varieties (Figure 4b) In addition, the analysis of 2AP in the

pANGmAMADH2-transformed calli compared with wild

type indicated that 2AP was synthesized in the

RNAi-trans-formed calli (Table 2)

An enzymatic assay of AMADH activity in PAGE gels

was performed to verify whether the product of

GmA-MADH2 was inhibited by RNAi The result clearly showed

that the candidate AMADH isozyme, which was thought

to be encoded by GmAMADH2, disappeared from the

extracts of calli from both RNAi lines The absence of the

AMADH band from the RNAi lines was similar to that in

the aromatic varieties, Chamame and Kaori hime, which

were used as controls (Figure 5) However, it is worth

not-ing that the band patterns of AMADH isozymes in young

seeds and calli were different A second isozyme band (band b, Figure 5) was also completely absent from the aromatic soybean calli in agreement with the pANGmA-MADH2-transformed calli, although this isozyme normally appeared in both aromatic and nonaromatic seeds Another isozyme band (band a, Figure 5) that was highly stained in young seeds of both aromatic and nonaromatic soybeans was faintly stained in the callus samples of all varieties

Plant BADH⁄ AMADH family BADH⁄ AMADH homologous protein sequences were retrieved from 11 flowering plant (Angiosperm) genomes,

(a)

(b)

Figure 4 (a) The structure of the pANGmA-MADH2 RNAi vector (b) RT-PCR analysis of endogenous GmAMADH2, the transgene and GmAMADH1, and RNA gel blot of the siRNA specific to GmAMADH2 among the Jack and CM60 wild-type varieties and the correspond-ing RNAi lines, Jack-RNAi and CM60-RNAi, respectively Lectin was used as a control for RT-PCR analysis rRNA was stained with 2% methylene blue as a loading control DNA oli-gomers of 22 and 24 nt were used as size m-arkers for siRNA.

Table 2 Contents of 2AP in soybean calli identified in wild type

and RNAi transgenic lines

Variety

2AP content (ppb; as FW) (mean ± SD) Phenotype

Wild type

Yuagari musume 457.3 ± 32.3 Aromatic

Kaori hime 325.3 ± 48.3 Aromatic

Chiang Mai 60, CM60 n.d Nonaromatic

Transgenic lines

CM60-RNAi (5-1) 324.2 ± 45.2 Aromatic

Jack-RNAi (2) 343.2 ± 50.2 Aromatic

n.d., not detected.

Figure 5 AMADH gel activity assay of the crude extracts from calli of the Jack and CM 60 wild-type varieties and the corresponding RNAi lines, Jack RNAi and CM 60 RNAi, respectively The varieties Cham-ame and Kaori hime were used as aromatic line controls for the enzyme activity in callus tissues The varieties Jack and Kaori hime were used as nonaromatic and aromatic controls, respectively, for the enzyme activity in seeds The candidate AMADH activity band is indi-cated by an arrow (c).

six dicot species and five monocot species, a lycophyte

genome, a chlorophyte genome and a bryophyte genome

In all flowering plants, two homologues of BADH⁄

AMA-DHs were identified from each species, whereas the three

primitive genomes contained only one BADH⁄ AMADH

The phylogenetic analysis showed that, among the

flower-ing plant genomes, two distinct groups were clearly

defined as dicot and monocot groups (Figure 6a) Among

the monocots, two complete orthologous subgroups

con-taining all five monocot species were clearly identified By

contrast, an orthologous group was not found among the

dicot species The two BADH⁄ AMADH homologues in

each dicot species were likely to be clustered within the same species However, in some closely related species, such as pea and soybean, orthologous groups were also clustered For Arabidopsis, one homologue was out-grouped from the other dicots

According to an in silico analysis of the NAD-dependent aldehyde dehydrogenase protein domain, an aldehyde dehydrogenase cysteine active site with the PROSITE regular expression [FYLVA] x {GVEP} {DILV} G [QE] -{LPYG} - C - [LIVMGSTANC] - [AGCN] - {HE} - [GSTAD-NEKR] was detected in all BADH⁄ AMADH homologues The whole set of BADH⁄ AMADH homologous proteins

(a)

(b)

Figure 6 (a) Phylogenetic tree of BADH ⁄

-AMADH homologues among the higher

plants Pisum sativum (CAC48392.2_PISSA

and CAC48393.1_PISSA), Zoysia tenuifolia

(BAD34953.1_ZOYTE and

BAD34949.1_ZO-YTE), Oryza sativa (Os04g0464200_ORYSA

and Os08g0424500_ORYSA), Zea mays

(ACF87737.1_ZEAMA and

N-P_001105781.1_ZEAMA), Sorghum bicolor

(Sb07g020650.1_SORBI and

Sb06-g019200.1_Sb06g019210.1_SORBI),

Arabid-opsis thaliana (At3g48170.1_ARATH and

At1g74920.1_ARATH), Populus trichocarpa

(661953_POPTR and 666405_POPTR), Glycine

max (BAG09376.1_GLYMA: GmAMADH2

and BAG09377.1_GLYMA: GmAMADH1),

Amaranthus hypochondriacus

(AAB58165.1_-AMAHY and AAB70010.1_(AAB58165.1_-AMAHY), Atriplex

hortensis (ABF72123.1_ATRHO and

P42757.-1_ATRHO) and Hordeum vulgare

(BA-B62846.1_HORVU and BAB62847.1_HORVU);

a lycophyte (Selaginella moellendorffii;

2439-4_CHLRE); a chlorophyte (Chlamydomonas

reinhardtii; 174224_SELMO); and a bryophyte

(Physcomitrella patens; EDQ78577.1_PHYPA).

The numbers indicated for each node are the

bootstrap values (b) Consensus sequences of

the aldehyde dehydrogenase cysteine active

site domains in BADH ⁄ AMADH homologous

proteins The BADH ⁄ AMADH homologues

were clustered as monocot, dicot and

primitive groups The two subgroups of the

monocot class are separated by a line The

cysteine active site in each sequence is

highlighted with a grey box The distinct

amino acids in the two subgroups of

monocot homologues are underlined The

PROSITE regular pattern of the aldehyde

dehydrogenase cysteine active site domain is

provided.

could be separated into two major groups according to

two consensus domain patterns, FANAGQVCSATS and

FWTNGQICSATS The former was found only in a

mono-cot subgroup The latter was found in another monomono-cot

subgroup as well as in most of the dicot BADH⁄ AMADHs,

except for soybean and pea, for which one homologue

contained FFTNGQICSATS In the primitive species that

have only one BADH⁄ AMADH protein, the BADH ⁄

AMA-DHs contained the second type of protein domain

(Figure 6b)

Discussion

Loss of AMADH from aromatic soybeans

2AP was detected by HSGC in all varieties of aromatic

soy-bean, confirming that it could be the potent aroma

com-ponent in soybeans as reported previously (Fushimi and

Masuda, 2001) Currently, the role of 2AP biosynthesis in

plants is unclear; however, 2AP could be formed to

detox-ify the 4-aminobutanal, which is a reactive amino-carbonyl

compound that accumulates when AMADH is inactivated

The accumulation of 4-aminobutanal has even been

shown in an Escherichia coli AMADH-deficient mutant

(Samsonova et al., 2005) According to the enzymatic

assay in this study, it is noteworthy that four enzymatic

activity bands were detected when 4-aminobutanal was

utilized in the AMADH enzymatic staining assay using

(NH4)2SO4-fractionated extracts from soybean seeds

(Fig-ure 1) This could suggest the existence of putative

AMADH isoforms or nonspecific aldehyde dehydrogenases

(ALDHs) that can metabolize 4-aminobutanal (Sˇebela

et al., 2001) The patterns of the AMADH activity bands in

the native PAGE assays for the extracts from young seeds

and calli of soybeans were not similar The reason

underly-ing this phenomenon remains unclear It is possible that

some of the AMADH isoforms were not present in calli or

that different sets of genes are expressed in calli and

seeds Nevertheless, the candidate AMADH enzymatic

band was consistently distinguishable between the

aro-matic and nonaroaro-matic soybeans in assays of both seed

and callus samples

GmAMADH1 and GmAMADH2 might have only

AMADH and not BADH function

In this study, we tested the two possible substrates,

beta-ine aldehyde and 4-aminobutanal, and found that only

4-aminobutanal was utilized by the enzymes extracted

from soybean seeds and calli Because the activity staining method used in this study was able to detect both BADH and AMADH activities, this result could suggest that there

is no BADH activity in soybean seed and callus extracts This may also imply that both GmAMADH1 and GmA-MADH2 encode enzymes that only function as AMADHs The lack of BADH activity has also been reported in pea (Pisum sativum) as the pea AMADH did not oxidize beta-ine aldehyde and elementary aldehydes (Sˇebela et al., 2000) On the contrary, in some monocots, such as rice (O sativa), barley (Hordeum vulgare) and oat (A Sativa), the Os2AP (BADH2) orthologues show very low BADH activity on betaine aldehyde, whereas proteins that are orthologous to BADH1 show moderate to high activity towards this substrate (Livingstone et al., 2002, 2003; Bradbury et al., 2008; Takashi et al., 2008) However, both paralogues, BADH1 and BADH2, in those monocots show a broad affinity for a range of amino aldehydes This might support our finding in this study that the BADH1 orthologues, which contain the monocot-specific consen-sus domain (Figure 6b), have been duplicated from the AMADH ancestor and then evolved the BADH function

GmAMADH2 is the key gene responsible for 2AP biosynthesis in soybeans

Both GmAMADH1 and GmAMADH2 could be thought of

as candidate genes for the aroma trait in soybeans owing

to their high similarities to rice Os2AP In this study, we clearly demonstrated that GmAMADH2 is the key gene associated with 2AP biosynthesis Both natural aromatic varieties, which contain an inactive form of GmAMADH2, and the GmAMADH2-RNAi knock-down lines lacked AMADH enzymatic activity, which probably resulted in the synthesis of 2AP Previous reports on the genetic control

of the aroma trait in vegetable soybean have shown seg-regation of aromatic and nonaromatic seeds in an F2 pop-ulation to be 1 : 3 This suggests that a single recessive gene could control the trait (AVRDC, 2003) It is possible that the gene regulating the trait is GmAMADH2 Hence,

a functional marker for molecular breeding for aroma in soybean could be designed based on the sequence varia-tion in GmAMADH2 Although GmAMADH2 is considered

to be the key player, a role for GmAMADH1 in 2AP bio-synthesis could not be ruled out Because GmAMADH1 and GmAMADH2 are highly similar to each other at both the nucleotide and protein levels, and because they are almost identical in the conserved domain for NAD-depen-dent aldehyde dehydrogenases, the similarity in enzymatic

function could be expected As multiple AMADH activity

bands are shown in Figures 1 and 5, one might think that

one of those bands is GmAMADH1 However, inactivation

of GmAMADH1 still needs to be performed to verify this

prediction In rice, another homologue of Os2AP, BADH1,

has been suggested to also play a role in 2AP biosynthesis

(Bradbury et al., 2008), but clear evidence has not yet

been reported

GmAMADH2 is independently mutated in soybean,

and its sequence variation is not related to those

found in rice Os2AP

Although the molecular mechanism that regulates 2AP

biosynthesis in soybeans might resemble that in rice, the

sequence variation in the CDS of GmAMADH2 that leads

to the loss of function in all aromatic soybean varieties is

not related to sequence variations previously reported in

rice Os2AP The sequence variations in Os2AP (Kovach

et al., 2009) and GmAMADH2 (this study) might have

evolved independently, but they appear to have the same

effect of inactivation of AMADH In rice, the major allele

of Os2AP in aromatic rice varieties is an 8-bp deletion in

exon 7 (Kovach et al., 2009) However, several aromatic

rice varieties contain other types of sequence variation in

Os2AP that cause the same effect Therefore, it could be

possible that this gene mutates easily This assumption is

possibly true in other plants and might explain why 2AP is

found in many plants Currently, only one type of

sequence variation, the TT deletion, has been identified

among the aromatic soybean varieties, as presented in this

study However, it is possible that other novel sequence

variations could exist

BADH⁄ AMADHs in dicots may have been recently

duplicated after the divergence of dicots and

monocots

As a result of the high similarity in the protein sequences,

BADH and AMADH are likely to have a common ancestor

Because only one BADH⁄ AMADH is found in primitive

spe-cies, the two homologues in flowering plants seem to

have evolved through a duplication of the ancestral gene,

which might have taken place after the divergence of

monocots and dicots Because two orthologous groups

can be clearly identified among the monocots, we predict

that the two BADH⁄ AMADHs in each monocot species

were duplicated in the monocot common ancestor before

the divergence of the monocot species By contrast, the

high similarity of the two BADH⁄ AMADHs in species of dicots indicates that the two genes might have been inde-pendently duplicated in each species after the divergence

of the dicot species The difference between the two sub-groups of BADH⁄ AMADH among the monocots revealed

by the two consensus sequences of the protein domain might explain the differences in the enzymatic function of the two BADH⁄ AMADH homologues in monocots, as pre-viously reported (Bradbury et al., 2008) According to the phylogenetic tree in this study, the true orthologues of the two BADH⁄ AMADH homologues can be distinguished among the monocots Therefore, proteins in the same orthologous group as rice Os2AP might be involved in 2AP biosynthesis However, orthology among dicots or comparison across dicots and monocots is not possible Therefore, both homologues of BADH⁄ AMADH in this gene family need to be analysed in dicots to identify the candidate gene for the aroma

In conclusion, aromatic soybeans share a similar mecha-nism with aromatic rice for 2AP biosynthesis, which involves the inactivation of AMADH Because this molecu-lar mechanism is conserved in distantly evolutionarily related species such as rice and soybean, we suggest that this regulation might also be shared in other plants These data can be used to engineer a metabolic pathway for 2AP production through inactivation of AMADH in other plants that do not accumulate 2AP

Experimental procedures Plant materials

Ten varieties of aromatic and nonaromatic vegetable soybean (G max L.) were used in this experiment The five aromatic soy-bean varieties (Chamame, Kouri, Kaori hime, Yuagari musume and Fukunari) and the three nonaromatic soybean varieties (Oku-hara Wase, Oishi Edamame and Shirono Mai) were collected from

a market in Japan Another nonaromatic soybean, Jack, was pro-vided by Professor Randall L Nelson, USDA, Agricultural Research Service, IL, USA The nonaromatic variety, Chiang Mai 60 (CM60), was provided by the Chiang Mai Field Crop Research Center, Department of Agriculture, Thailand Mature seeds of each variety were grown in pots under open-air conditions at Kasetsart Univer-sity, Nakhon Pathom, Thailand Fresh pods from each plant were harvested between October and November 2008.

2AP analysis in soybean seeds

Automated HSGC using an Agilent Technologies (Wilmington, DE, USA) model 6890N gas chromatograph, an Agilent Technologies model G1888 headspace autosampler and a nitrogen–phosphorus

detector (NPD) was performed for headspace volatile extraction

and quantitative analysis of 2AP in the soybean samples The

HSGC conditions were the same as those reported by Sriseadka

et al., 2006 with some modifications Chromatographic separation

was performed on a fused silica capillary column, phase HP-5MS

(60 m · 0.32 mm i.d · 1.0 lm) (Agilent Technologies) Purified

helium gas at a flow rate of 1 mL ⁄ min was used as the GC carrier

gas 2AP contents in soybean samples were determined using an

internal standardization method with 2,4-dimethylpyridine

(2,4-DMP) as the internal standard Identification of 2AP in the

soy-bean headspace was accomplished using an HP 5973

mass-selec-tive detector (Agilent Technologies, Palo Alto, CA, USA) equipped

with the Wiley 7N Mass Spectral Library and the NIST 05 Mass

Spectral Library, both purchased from Agilent Technologies

Addi-tionally, structural confirmation by mass spectral comparison with

2AP was performed The mass spectrometer was operated in the

electron impact mode with an electron energy of 70 eV, ion

source temperature of 230 C, quadrupole temperature of

150 C, mass range m ⁄ z of 29–550, scan rate of 0.68 s ⁄ scan and

EM voltage of 1423 V The GC–MS transfer line was set to

280 C The system operation, as well as data acquisition,

collec-tion and evaluacollec-tion, was accomplished using Agilent ChemStacollec-tion

software version A.01.04 and B.01.03 (Agilent Technologies,

Waldbronn, Germany).

Crude and fractionated extract preparation

All extraction procedures were performed at 4 C Young soybean

seeds and callus tissue were ground into fine powder, and then

crude enzymes were extracted with cold extraction buffer

(100 m M KPi (pH 7.5), 10% (v ⁄ v) glycerol, 2% (w ⁄ v) PVPP, 1 m M

EDTA, 1 m M NAD and 1 m M DTT, at a ratio of 400 mg fresh

weight per mL) The homogenate was centrifuged at 20 800 g

for 15 min The supernatant was fractionated with solid

(NH 4 ) 2 SO 4 , and the fraction containing 55%–75% saturation was

collected The fraction was resuspended in 500 lL of 100 m M KPi

(pH 7.5) buffer and desalted through a NAP-5 column (GE

Health-care Biosciences, Uppsala, Sweden) A solution of 1 m M NAD,

1 m M EDTA and 1 m M DTT was added to obtain desalted

solu-tions The crude extracts were assayed immediately.

AMADH activity gel staining assay

A nondenaturing polyacrylamide mini-gel system (Bio-Rad

Labo-ratories, Inc., Hercules, CA, USA) was used in this study to assay

the AMADH activity Total enzyme extracts (20 lg in 16 lL)

were mixed with 4 lL of gel-loading buffer containing 50%

(v ⁄ v) glycerol and 0.05% (w ⁄ v) bromphenol blue The samples

were then separated on 10% PAGE gels, Tris–HCl, pH 8.8 The

electrode buffer contained 25 m M Tris–HCl and 192 m M glycine.

Separation was performed with constant current at 20 mA ⁄ gel

at 4 C Staining was performed at 37 C in a solution

contain-ing 100 m M glycine–NaOH buffer (pH 9.5), 1 m M NAD+, 5 m M

4-aminobutanal or betaine aldehyde, 1 m M

(3-4,5-dim-ethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) and

0.15 m M 1-methoxy-phenazine methosulphate until the bands

were visible The reaction was terminated by adding 10% acetic

acid and then distilled water.

Reverse transcriptase polymerase chain reaction

Total RNA was isolated from immature seeds, leaves and calli of the ten varieties using an RNeasy Plant mini kit (Qiagen, Valencia,

CA, USA) The DNAase-treated RNA was reverse-transcribed and PCR-amplified using Titan One Tube RT-PCR kit (Roche Applied Science, Penzberg, Germany) according to the manufacturer’s instructions The reverse transcription and PCR thermal cycle con-ditions were performed as follows: reverse transcription at 50 C for 30 min, then PCR with initial heating at 95 C for 3 min fol-lowed by 27–37 cycles of 95 C, 30 s; 55 C, 30 s; 68 C, 1 min and a final extension at 68 C for 5 min The number of cycles was adjusted to avoid over-cycling, and all RT-PCR assays were carried out in triplicate GUS linker was used to determine the lev-els of transcription from the RNAi construct, and Lectin was used

as control to determine the mRNA amount The primers for GmA-MADH1 were 5¢-TGAAGCTGGTGCTCCTTTGT-3¢ and 5 AAGATGGTCCATTCAGCAGT-3¢ The primers for GmAMADH2 were 5¢-TGAAGCGGGTGCTCCTTTAG-3¢ and 5¢-AATATGGTCC-ATTCAGCAGC-3¢ The primers for GUS linker were 5¢-CAT-GAAGATGCGGACTTACG-3¢ and 5¢-ATCCACGCCGTATTCGG-3¢ The primers for Lectin were 5¢-TCAACGAAAACGAGTCTGGTG-3¢ and 5¢-GGTGGAGGCATCATAGGTAAT-3¢.

Polymerase chain reaction (PCR)

Genomic DNA was extracted from young leaves using the DNeasy Plant mini kit (Qiagen) PCR was performed in 25 lL reaction mix-tures containing 50 ng of genomic DNA template, 0.1 m M of dNTPs, 0.25 m M of each forward and reverse primer, 0.25 units

of Taq DNA polymerase, 2.0 m M MgCl2 and 1· Thermophilic DNA Polymerase buffer (Promega, Madison, WI, USA) After being preheated at 94 C for 2 min, the PCR was carried out for 30 cycles under the following conditions: 94 C denaturation for

30 s, 55 C annealing for 30 s and a 72 C extension for 1 min, with a final extension at 72 C for 5 min Primers used for PCR are listed in Supplementary Table I.

DNA sequencing and sequence assembly

The amplified PCR fragments were purified and cloned into the pGEM-T Easy Vector (Promega) The templates were sequenced in both directions with an automatic sequencer using the ABI PRISM Big Dye Terminator Cycle (Applied Biosystem ⁄ Perkin-Elmer, San Jose, CA, USA) Sequences were assembled and viewed using the phred ⁄ phrap ⁄ consed software (http:// www.phrap.org).

Construction of the RNAi vector

The GmAMADH2 fragment was amplified from the genomic DNA

of the Jack variety by PCR using the forward primer 5¢-CAC-CATGAGCATCCCAATTCCCCA-3¢ and the reverse primer 5¢-TT-CGAGTTTTGCTAGTTCAGG-3¢ The amplified PCR fragment was cloned into the Gateway pENTR ⁄ D-TOPO cloning vector (Invitrogen, Carlsbad, CA, USA), which carries two recombination sites (attL1 and attL2) for the LR Clonase reaction Subsequently, the target