Legume Transcription Factors: Global Regulators of Plant Development and Response to the Environment^,

Transcription factors (TFs) are DNA-binding proteins that interact with other transcriptional regulators, including chromatin remodeling/modifying proteins, to recruit or block access of RNA polymerases to the DNA template. Plant genomes devote approximately 7% of their coding sequence to TFs, which is a testament to the complexity of transcriptional regulation in these organisms. Extensive sequencing of cDNA and genomic DNA indicates that legumes encode upwards of 2,000 TFs per genome. Less than 1% of these have been characterized genetically, although TFs likely played seminal roles in legume evolution and clearly now play crucial roles in plant development and differentiation. Here we review the literature on legume TFs and describe technological developments that are paving the way for rapid and systematic characterization of TFs and the genetic regulatory networks they control.

Plants are amazing organisms. Not only are they able to build complex organic superstructures from simple inorganic molecules that ensure their growth and reproductive success, but they do this while fixed in space and subject to environmental extremes of light, temperature, water, and nutrients, and to biological challenges from competitors, pests, and pathogens. Evolution has endowed plants with a flexible developmental program that enables them to elaborate new vegetative organs and attune reproduction to prevailing environmental conditions. Plant cells can also differentiate in the short term to cope with more immediate environmental challenges. Plant development and differentiation are programmed primarily at the level of gene transcription, which is controlled by TFs and other proteins that either recruit or block access of RNA polymerases to the DNA template. TFs are usually defined as sequence-specific DNA-binding proteins that are capable of activating and/or repressing transcription. Plant genomes appear to encode many more TFs than those of animals, such as Caenorhabditis elegans and Drosophila melanogaster, which indicates that transcriptional regulation in plants is at least as complex as in animals (Riechmann et al., 2000). Arabidopsis (Arabidopsis thaliana) possesses upwards of 1,800 TF genes representing more than 7% of all protein-coding genes (Riechmann et al., 2000; Guo et al., 2005; Iida et al., 2005). Surprisingly, only one-tenth of these have been characterized genetically (Qu and Zhu, 2006) despite the enormous resources that have been devoted to Arabidopsis research over the past decade. Not surprisingly, we know far less about the role of TFs in other plant species. For instance, less than 1% of TF genes in the model legumes Lotus japonicus (or simply Lotus) and Medicago truncatula (or Medicago) have been genetically characterized. This makes review of the literature on legume TFs a relatively simple task at present, although there are signs that this situation will change rapidly over the next few years. First of all, it is already apparent that TFs play crucial roles in agriculturally important processes in legumes, such as symbiotic nitrogen fixation (SNF), so there is great incentive to learn more about this important class of regulatory proteins. Second, the genome of three legume species, Medicago, Lotus, and soybean (Glycine max), will be completed or largely so in the next 2 years, which will enable the identification of most of the TFs in these species via bioinformatics approaches. Finally, numerous tools for functional genomics have been and are being developed that will facilitate rapid and systematic functional characterization of large numbers of TFs. This review summarizes our current state of knowledge about legume TFs and considers the opportunities and challenges for continued research in this area.

THE DYNAMIC TRANSCRIPTOME

As a backdrop to our discussion of legume TFs, it is salient that recent transcriptomic studies, using arrays of cDNA or oligonucleotides to measure transcript levels, have identified thousands of legume genes that are differentially expressed during various types of plant-microbe interactions (Colebatch et al., 2002, 2004; Liu et al., 2003; Barnett et al., 2004; El-Yahyaoui et al., 2004; Kouchi et al., 2004; Lee et al., 2004; Manthey et al., 2004; Mitra et al., 2004; Moy et al., 2004; Suganuma et al., 2004; Hohnjec et al., 2005; Zou et al., 2005; Alkharouf et al., 2006; Lohar et al., 2006; Starker et al., 2006; Zabala et al., 2006), development and differentiation (Vodkin et al., 2004; Aziz et al., 2005; Firnhaber et al., 2005; Dhaubhadel et al., 2007), and in response to abiotic stress (Ainsworth et al., 2006; Buitink et al., 2006). Invariably, TFs have been found among differentially expressed genes, implicating them in the regulation of specific developmental processes or responses to the biotic and abiotic environment. Such guilt by association is a theme that is elaborated upon below.

IDENTIFICATION OF PUTATIVE TFs

Bioinformatics approaches have been instrumental in identifying putative TF genes in plants. TF families are generally defined by the types of DNA-binding domain contained by proteins in the family (Table I; Fig. 1) and putative TF genes have been identified primarily on the basis of DNA sequences within the gene that encode known DNA-binding domains (Riechmann et al., 2000; Guo et al., 2005; Iida et al., 2005). BLAST and similar searches that look for extended sequence homology between query sequences and known TFs have also been used to identify putative TFs (Iida et al., 2005). BLAST searches that utilize well-curated protein databases, such as UniProt (http://www.expasy.uniprot.org/database/knowledgebase.shtml), can also be used to support TF annotations that were made initially on the basis of the presence of a DNA-binding or other characteristic domain. We searched the current International Medicago Gene Annotation Group (IMGAG) dataset, which contains 40,568 predicted proteins, for the presence of sequences encoding DNA-binding and other TF domains to identify putative TF genes in this species (Supplemental Table S1). A subset of these TFs, obtained from an earlier release of IMGAG gene annotations, was verified by BLAST analysis, which resulted in a list of 1,084 putative TF genes (Table I). We have designed and tested gene-specific primers for each of these for use in high-throughput quantitative reverse transcription (qRT)-PCR analysis and have plans to develop this resource further to facilitate transcript analysis of all Medicago TF genes in the future (K. Kakar and M.K. Udvardi, unpublished data).

Table I.

Classification of putative TFs of Medicago into families and subfamilies

IMGAG proteins were classified as putative TFs if they contained characteristic DNA-binding or other characteristic TF domains and if annotations of matching proteins obtained by BLAST searches were consistent with such a classification. TF families previously identified in plants are presented in the first part of the table, whereas potentially novel plant TF families, which were identified by the presence of domains associated with TFs and other transcriptional regulators outside the plant kingdom, are presented in the latter part of the table. Plant-specific TF families and subfamilies are indicated in bold (according to Riechmann, 2002). D, DNA-binding domain; P, protein-protein interaction domain; NA, nucleic acid (DNA and RNA) binding domain; RD, receiver domain; LBD, ligand binding; TA, transcriptional coactivator.

TF Family	No. of Genes	Characteristic Domain (InterPro No.)	Domain Function	Domain Description
MYB/HD like	77	IPR001005; IPR009057	D	Myb, DNA binding; homeodomain like
MYB	59	IPR001005	D	Myb, DNA binding
C2H2 (Zn)	64	IPR007087	NA	Zn-finger, C2H2 type
AP2/EREBP	55	IPR001471	D	Pathogenesis-related transcriptional factor and ethylene response factor
bHLH	50	IPR001092	D	Basic helix-loop-helix dimerization region bHLH
HD like	50	IPR009057	D	Homeodomain like
HD family		IPR001356	D	Homeobox
HD	25
HD-ZIP	5	IPR006712	P	HD-ZIP protein, N terminus
HD-PHD-finger	2	IPR001965	P	Zn-finger like, PHD-finger
MADS	48	IPR002100	D	TF, MADS-box
bZIP	42	IPR004827	D	Basic Leu zipper (bZIP) TF
PHD	34	IPR001965	P	Zn-finger like, PHD-finger
WRKY family		IPR003657	D	DNA-binding WRKY
WRKY	29
LLR WRKY	1	IPR001611		Leu-rich repeat
ABI3/VP1	29	IPR003340	D	TF B3
NAC	29	IPR003441	D	No apical meristem (NAM) protein
C3H-type 1 (Zn)	27	IPR000571	D	Zn-finger, C-x8-C-x5-C-x3-H type
ARF	23	IPR003340, IPR010525, IPR011525	D
JUMONJI	20	IPR003347	D	TF jumonji, jmjC
GRAS	19	IPR005202	P	GRAS TF
HMG	15	IPR000637	D	HMG-I and HMG-Y, DNA binding
AS2	14	IPR004883	P	Lateral organ boundaries
C2C2 (Zn)
Dof	14	IPR003851	D	Zn-finger, Dof type
GATA	7	IPR000679	D	Zn-finger, GATA type
CO like	6	IPR000315	D	Zn-finger, B-box
YABBY	5	IPR006780	D	YABBY protein
CCAAT-HAP3 type	12	IPR003958	D	TF CBF/NF-Y/archaeal histone
GRF	8	IPR010666	D	Zn-finger, GRF type
SBP	8	IPR004333	D	SBP
EIL	7	IPR006957	D	Ethylene insensitive 3
LIM	7	IPR001781	P	Zn-binding protein, LIM
SNF2	6	IPR000330	D	SNF2 family N-terminal domain
E2F/DP	5	IPR003316	D	TF E2F/dimerization partner (TDP)
TCP	5	IPR005333	D	TCP TF
FHA	5	IPR000253	D	Forkhead associated
ARID	4	IPR001606	D	AT-rich interaction region
HSF	4	IPR000232	D	Heat shock factor (HSF)-type, DNA binding
AUX/IAA	3	IPR003311	D	AUX/IAA protein
SRS	3	IPR006510	D	Zn-finger, LRP1 type
TUB	3	IPR000007	D	Tubby
ZIM	3	IPR010399	D	ZIM
DDT	3	IPR004022	D	DDT
ZF-HD	2	IPR006455	D	Homeobox domain, ZF-HD class
MBF1	2	IPR001387	D	Helix-turn-helix type 3
S1Fa like	2	IPR006779	D	DNA-binding protein S1FA
CAMTA	2	IPR005559	D	CG-1
LFY	1	IPR002910	D	Floricaula/leafy protein
NIN like	1	IPR003035	D	Plant regulator RWP-RK
TAZ	1	IPR000197	P	Zn-finger, TAZ type
Potentially novel plant TFs and transcriptional regulators
CCHC (Zn)	112	IPR001878	NA	Zn-finger, CCHC type
RR	16	IPR001789, IPR011006	RD	Response regulator receiver
DHHC (Zn)	14	IPR001594	D or P	Zn-finger, DHHC type
HTH
FIS	11	IPR002197	D	Helix-turn-helix, Fis type
AraC	2	IPR000005	D	Helix-turn-helix, AraC type
BTB/POZ	7	IPR000210	P	BTB
TTF-type (Zn)	6	IPR006580	D	Zn-finger, TTF type
BD	6	IPR001487	P	Bromodomain
λ-DB	3	IPR010982	D	λ_DNA_bd
TrpR	3	IPR010921	D	Trp repressor/replication initiator
TPR	3	IPR001440	P	Tetratricopeptide TPR_1
KRAB-box	2	IPR001909	P	KRAB box
NRs	2	IPR008946	LBD	Steroid nuclear receptor, ligand binding
R3H	2	IPR001374	NA	Single-stranded nucleic acid binding R3H
YEATS	2	IPR005033	TA	YEATS
U1-type (Zn)	2	IPR003604	NA	Zn-finger, U1 type
A20 like	2	IPR002653	P	Zn-finger, A20 type
Euk_TF	1	IPR008917	D	Euk_TF_DNA_bd
NGN	1	IPR006645	D	NGN
p53 like	1	IPR008967	D	p53-like TF, DNA binding
SSB protein	1	IPR011344	D	Single-strand binding protein
ssDB TR	1	IPR009044	D	Single-strand DNA-binding transcriptional regulator
TCoAp15	1	IPR003173	D	Transcriptional coactivator p15
BED-type (Zn)	1	IPR003656	D	Zn-finger, BED-type predicted
TCoA	1	IPR009255	TA	Transcriptional coactivation
Tc/PD	1	IPR001533	TA	Transcriptional coactivator

Figure 1.

Relationships and domain shuffling between Medicago TF families. TF families are represented by circles whose size is proportional to the number of members in the family. Domains are represented by rectangles, whose size is proportional to the domain length. DNA-binding domains appear in color. Protein-binding and other domains are hatched. Dashed lines indicate that a given domain is characteristic of the family or subfamily to which it is attached. Dotted lines indicate domains that define a potentially novel TF family or subfamily. Based on Figure 1 of Riechmann (2002).

FUNCTIONAL CHARACTERIZATION OF TFs

As noted above, very few legume TFs have been characterized genetically so far (Table II). An important feature of legumes that sets them apart from plants in other families is their ability to form nitrogen-fixing symbioses with soil bacteria, called rhizobia. These bacteria take up intracellular residence in specialized organs, called nodules, that develop on roots and stems specifically for the purpose of SNF. Given the importance of SNF to sustainable agriculture, it has been a major focus of legume research over the past few decades and one of the few areas of legume biology where the role of TFs has been firmly established. The first TF gene implicated in SNF was Nin, for nodule inception, which was cloned from a transposon-tagged mutant of Lotus that was unable to form nodules (Schauser et al., 1999). NIN was the founding member of a novel family of putative TFs in higher plants, now called the NIN-like family (Fig. 1), and shares homology with Chlamydomonas minus dominance proteins, which are developmental regulators in these algae. Nin-like genes are widespread in the plant kingdom. However, their predicted DNA-binding and gene regulation activities are yet to be proven formally. Classical or forward-genetics approaches have subsequently identified three other TF genes in Lotus (Nishimura et al., 2002) or Medicago (Kaló et al., 2005; Smit et al., 2005) that are essential for nodule development. NODULATION SIGNALING PATHWAY1 (NSP1) and NSP2 of Medicago are both GRAS-family proteins, putative TFs that transduce the bacterial Nod factor signal and induce expression of plant nodulin genes, which are presumably required for nodule development (Kaló et al., 2005; Smit et al., 2005). Both NSP1 and NSP2 were isolated by map-based cloning. Orthologs of NSP1 and NSP2 were subsequently isolated from Lotus by candidate gene approaches (Heckmann et al., 2006) and positional cloning (Murakami et al., 2007). The Lotus LjBzf gene encodes a bZIP TF that negatively regulates nodule development (Nishimura et al., 2002). The bzf/sym77 mutant exhibits not only faster and more prolific nodulation, but also light and gravity response defects reminiscent of the hy5 mutant of Arabidopsis. A Lotus homolog (LjBzf) of the Arabidopsis HY5 gene was subsequently cloned and found to cosegregate with the mutant phenotype. The bzf/sym77 mutant contained a single base pair mutation in a splice donor site of this gene. Finally, a wild-type version of LjBzf complemented the mutant phenotype in transgenic plants, confirming its role in the regulation of nodule development. All four of the genes mentioned above have homologs in nonlegume species, such as Arabidopsis, which suggests that they have been recruited rather than invented during evolution to fulfill roles in nodule development (Szczyglowski and Amyot, 2003). This certainly seems to be the case for LjBzf and Mszpt2-1 (see below) given the additional nonsymbiotic phenotypes of these mutants (Frugier et al., 2000; Nishimura et al., 2002). Interestingly, the nonsymbiotic phenotypes of the Lotus bzf/sym77 and Arabidopsis hy5 mutants are not identical (the latter shows enhanced lateral root initiation, whereas the former does not), reflecting evolutionary divergence in gene function in the two plant lineages quite apart from the acquisition of the novel symbiotic function in the legume lineage. This theme is reiterated below for TFs involved in floral development.

Table II.

Genetically characterized TFs in legumes

TF Name	TF Family	Process Regulated	Species	Method	References
NIN	NIN like	Nodule development	Lotus; pea	Transposon mutagenesis	Schauser et al. (1999); Borisov et al. (2003)
LjBzf	bZIP	Nodule development	Lotus	Positional cloning	Nishimura et al. (2002)
NSP1	GRAS	Nodule development	Medicago; Lotus	Positional cloning	Smit et al. (2005); Heckmann et al. (2006)
NSP2	GRAS	Nodule development	Medicago; Lotus	Positional cloning	Kaló et al. (2005); Heckmann et al. (2006)
Mszpt2-1	C2H2 (Zn)	Nodule development	Medicago	Antisense	Frugier et al. (2000)
LjNDX1, LjNDX2	HD	Nodule function	Lotus	Antisense	Gronlund et al. (2003)
MtHAP2-1	CCAAT binding	Nodule development	Medicago	RNAi	Combier et al. (2006)
UNI/LjFLO	LFY	Flower and leaf development	Pea; Lotus	Candidate gene approach	Hofer et al. (1997); Dong et al. (2005)
PIM, MtPIM	MADS	Floral meristem identity	Pea; Medicago	Candidate gene approach, transposon mutagenesis	Berbel et al. (2001); Taylor et al. (2002); Benlloch et al. (2006)
PsPi	MADS	Floral development	Pea	Complementation	Berbel et al. (2005)
LjCYC2	TCP	Floral development	Lotus	Candidate gene approach	Feng et al. (2006)
PHANTASTICA	MYB	Compound leaf development	Pea	Candidate gene approach	Tattersall et al. (2005)
PvNAP	NAC	Leaf senescence	Common bean	Ectopic overexpression	Guo and Gan (2006)
WXP1, WXP2	AP2 domain	Wax biosynthesis	Medicago	Ectopic overexpression	Zhang et al. (2005); Zhang et al. (2007)
Mszpt2-1	Kruppel like	Salt tolerance	Medicago	Antisense	Merchan et al. (2003)
CAP2	AP2	Salt, drought tolerance, growth development	Chickpea (Cicer arietinum)	Ectopic overexpression	Shukla et al. (2006)
SCOF-1	C2H2 (Zn)	Cold tolerance	Soybean	Ectopic overexpression	Kim et al. (2001)
Alfin1	Zn-finger	Salt tolerance, growth development	Alfalfa	Overexpression, antisense	Winicov (2000)

Many TF genes have been found to be expressed during nodule development and differentiation (see above) and several groups are now using the tools of reverse genetics to decipher the roles of such genes in SNF. Three TFs have been implicated in nodule development or function in this way (Table II). The first of these was Mszpt2-1, a Kruppel-like TF of the C2H2 (Zn) family that was found to be essential for differentiation of the nitrogen-fixing zone of alfalfa (Medicago sativa) nodules via an antisense RNA approach (Frugier et al., 2000). A similar approach implicated the Lotus ndx gene family in nodule function and maintenance (Gronlund et al., 2003). Most recently, RNA interference (RNAi) revealed a key role in nodule development for MtHAP2-1, a member of the CCAAT-binding family of TFs (Combier et al., 2006). Interestingly, MtHAP2-1 was found to be regulated by microRNA169, revealing an important role for microRNA in the regulation of legume development.

Whereas research to identify TFs involved in SNF has profited little from previous work in nonlegumes such as Arabidopsis, knowledge from nonlegume models has been instrumental in identifying a number of TF genes involved in common plant processes, such as flower and leaf development (Table II). In fact, the first legume TF gene to be characterized functionally was pea (Pisum sativum) FLO, which was isolated by virtue of its sequence homology to the TFs FLO and LFY of snapdragon (Antirrhinum majus) and Arabidopsis, respectively. FLO and LFY control floral development in snapdragon and Arabidopsis, and a defect in pea FLO was subsequently found to be responsible for aberrant floral and leaf development in the pea unifoliata (uni) mutant (Hofer et al., 1997). The Lotus ortholog of FLO was later identified in the same way (Dong et al., 2005). Similar approaches were used to assign functions for PIM (a MADS family TF) in pea floral meristem determination (Taylor et al., 2002) and for PHANTASTICA (a MYB TF) in pea compound leaf development (Tattersall et al., 2005). Interestingly, Lotus has a duplicate pair of PHANTASTICA-like genes that probably have divergent functions in compound leaf development (Luo et al., 2005).

The value of computational approaches in identifying genes likely to be involved in various aspects of flowering was nicely illustrated by Hecht et al. (2005), who utilized sequence information from Arabidopsis TFs to identify homologs in model legume sequence databases, which were then used to design PCR-cloning strategies to isolate homologs from pea. The majority of Arabidopsis flowering genes were represented in pea and other legume sequence databases. However, several gene families, including the MADS-box, CONSTANS, and FLOWERING LOCUS T/TERMINAL FLOWER1 families, appeared to have undergone differential expansion, whereas other genes important in Arabidopsis, including FRIGIDA and members of the FLOWERING LOCUS C clade, were conspicuously absent from legumes. Several pea and Medicago orthologs mapped to syntenic chromosomal positions, demonstrating the benefit of parallel model systems for understanding flowering phenology in crop and model legume species.

TFs of the TCP family, named after the founding members TB1, CYC, and PCF, help to establish the pattern of flower petals (Cubas, 2004), which gives legume flowers their typical bilateral symmetry. Citerne et al. (2003) sequenced a number of CYC homologs, sorted them into clades by phylogenetic analysis, and discussed the difficulties of assigning orthologs in cross-species comparisons. Thus, genetic map position, mutant phenotypes, and/or complementation of Arabidopsis mutants have been used in addition to sequence similarity to infer orthology. The squared standard mutant of Lotus is defective in the TCP gene LjCYC2, which is required to establish floral bilateral symmetry and was cloned by a candidate gene/sequence homology approach (Feng et al., 2006). Adaxial expression of two CYC genes was observed in the developing floral meristem of lupin (Lupinus albus), which also has bilaterally symmetrical flowers (Citerne et al., 2006). Interestingly, evolution of radially symmetrical flowers in Cadia, which belongs to the same subclade as Lupinus, may have resulted from an expanded domain of expression of an orthologous CYC gene in the former (Citerne et al., 2006).

Cross-species complementation studies have indicated possible roles for several legume TFs. For example, Berbel et al. (2001) rescued the Arabidopsis apetala1 (ap1) floral development mutant using a PIM cDNA clone, which they called PEAM4. Later, the same group characterized PsPI, a pea MADS-box gene homologous to the petal and stamen identity genes PISTILLATA (PI), from Arabidopsis and GLOBOSA, from snapdragon. Interestingly, constitutive expression of PsPI in Arabidopsis rescued the floral defects caused by the strong pi-1 mutant allele, despite the fact that the pea protein, PsPI, lacked a particular C-terminal motif (Berbel et al., 2005). Similarly, Guo and Gan (2006) showed that overexpression of PvNAP, a kidney bean (Phaseolus vulgaris) NAC TF homologous to Arabidopsis AtNAP, successfully complemented the leaf abscission phenotype of an atnap null mutant, indicating a possible role of NAP in bean leaf abscission.

Finally, five legume TFs have been implicated in abiotic stress tolerance (Table II). One of these, alfalfa Mszpt2-1, which was mentioned previously in the context of nodule development, was found to be induced in roots by salt treatment. Inhibition of Mszpt2-1 by antisense RNA resulted in increased sensitivity of transgenic plants to salinity (Merchan et al., 2003). Overexpression of CAP2 and Alfin1 TFs in transgenic plants conferred salt tolerance and increased growth (Winicov, 2000; Shukla et al., 2006). Constitutive overexpression of SCOF-1, a soybean protein, increased cold tolerance of transgenic Arabidopsis and tobacco (Nicotiana tabacum) plants (Kim et al., 2001). Another example of successful leveraging of knowledge from nonlegumes for legume research is provided by the Medicago WXP1 gene, a member of the AP2/ethylene-responsive element-binding protein (EREBP) family of TFs. Several members of this family have been implicated in drought tolerance in Arabidopsis and other nonlegumes. Overexpression of Medicago WXP1 in alfalfa resulted in enhanced tolerance to drought stress, which correlated with increased wax deposition in the leaf cuticle (Zhang et al., 2005).

Whereas work on deciphering the roles of legume TFs is just beginning, considerable efforts have already been made to demonstrate the functionality of such proteins in terms of their DNA-binding and transactivation abilities and subcellular localization (Table III). Approaches for isolating legume TF genes have varied widely. Homology-based methods using TF DNA from other plant families have been successfully employed to identify specific classes of legume TFs. For instance, GmEREBP1 was isolated from a soybean root cDNA library screened with a probe that was PCR amplified using degenerate primers matching the conserved EREBP-coding domain (Mazarei et al., 2002). Three soybean DRE-binding proteins were identified by sequence homology to the AP2/EREBP consensus sequence via a BLAST search of the soybean EST database (Li et al., 2005). Several groups have used degenerate primers matching conserved TF domains for PCR amplification of legume TF sequences (Chern et al., 1996a, 1996b; Heard et al., 1997; Uimari and Strommer, 1997; Zucchero et al., 2001; Tucker et al., 2002).

Table III.

Biochemical and molecular characterization of legume TFs

NL, Nuclear localization. All other abbreviations are defined in the text.

TF Name	TF Family	Proposed Role	Species	Method	References
Ph_acut_ AY026054	bZIP	Abiotic stress	Phaseolus acutifolius	NL	Rodriguez-Uribe and O'Connell (2006)
Ph_vulg_ AF350505	bZIP	Abiotic stress	Kidney bean	NL	Rodriguez-Uribe and O'Connell (2006)
GmDREBa	AP2/EREBP	Abiotic stress	Soybean	Yeast one hybrid	Li et al. (2005)
GmDREBb	AP2/EREBP	Abiotic stress	Soybean	Yeast one hybrid	Li et al. (2005)
GmDREBc	AP2/EREBP	Abiotic stress	Soybean	Yeast one hybrid	Li et al. (2005)
SCOF-1	C2H2 (Zn)	Abiotic stress	Soybean	NL, yeast two hybrid	Kim et al. (2001)
CAP2	AP2/EREBP	Abiotic stress and development	Chickpea	EMSA, yeast one hybrid, NL	Shukla et al. (2006)
Alfin1	Alfin-like/PHD-finger	Abiotic stress and development	Alfalfa	EMSA	Bastola et al. (1998)
PLATZ1	PLATZ (Zn)	Cell division	Pea	EMSA	Nagano et al. (2001)
GmHZ1	HD-ZIP	Defense	Soybean	EMSA, NL	Wang et al. (2005)
SGBF-1	GBF	Development	Soybean	EMSA	Hong et al. (1995)
SGBF-2	GBF	Development	Soybean	EMSA	Hong et al. (1995)
STF1	bZIP with RING Zn-finger motif	Development	Soybean	EMSA	Cheong et al. (1998)
STF2	bZIP with RING Zn-finger motif	Development	Soybean	EMSA	Cheong et al. (1998)
STGA1	TGA-type bZIP	Development	Soybean	EMSA	Cheong et al. (1994)
PvTGA1.1	TGA-type bZIP	Leaf abscission	Kidney bean	EMSA	Tucker et al. (2002)
PvTGA2.1	TGA-type bZIP	Leaf abscission	Kidney bean	EMSA	Tucker et al. (2002)
PvTGA2.2	TGA-type bZIP	Leaf abscission	Kidney bean	EMSA	Tucker et al. (2002)
Myb26	MYB	Flower development	Pea	EMSA	Uimari and Strommer (1997)
ngl9	MADS box	Nodule and flower development	Alfalfa	EMSA	Zucchero et al. (2001)
nmhc5	MADS box	Nodule development	Alfalfa	EMSA	Heard et al. (1997)
GBP	GAGA-binding protein	Nodule function	Soybean	EMSA, yeast one hybrid	Sangwan and O'Brian (2002)
G/HBF-1	bZIP	Pathogen defense response	Soybean	EMSA	Dröge-Laser et al. (1997)
KAP-2	H-box binding	Phenylpropanoid biosynthesis	Kidney bean and Medicago	EMSA, in vitro transcription assay	Lindsay et al. (2002)
GmHdl56	HD-ZIP	Phosphate responses	Soybean	EMSA, DNase-I footprinting	Tang et al. (2001)
GmHdl57	HD-ZIP	Phosphate responses	Soybean	EMSA	Tang et al. (2001)
PvALF	ABI3 like	Seed development	Kidney bean	EMSA, transient expression assay, transactivation in planta	Bobb et al. (1997); Nag et al. (2005)
ROM1	bZIP	Seed development	Kidney bean	Transient expression assay, EMSA, DNase-I footprinting	Chern et al. (1996b)
ROM2	bZIP	Seed development	Kidney bean	Transient expression assay, EMSA, DNase-I footprinting	Chern et al. (1996a)
TGA1a	TGA-type bZIP	Seed development	Pea	EMSA, DNase-I footprinting, methyl interference assay	de Pater et al. (1994)
GmGT-2	Trihelix	Light responses	Soybean	EMSA	O'Grady et al. (2001)
PCF1	HMG	Photosynthesis	Pea	Filter-binding assay, DNase-I footprinting	Pwee et al. (1994); Webster et al. (1997)
VR-EIL1/2	EIL	Ethylene-signaling pathway	Vigna radiata	EMSA, NL, transient expression, transactivation in yeast	Lee and Kim (2003)
GmEREBP1	AP2/EREBP	Wounding and pathogen response	Soybean	EMSA	Mazarei et al. (2002)

Other TFs have been identified based on their ability to interact with known gene cis-elements. The use of cDNA expression libraries has been valuable in this regard. Two HD-ZIP proteins (GmHDL56/57) were identified using a 160-bp fragment of the VspB promoter (Tang et al., 2001). Similarly, G/HBF-1, SGBF-1/2, and GmGT-2 were isolated based on their ability to bind specific promoter elements (Hong et al., 1995; Dröge-Laser et al., 1997; O'Grady et al., 2001). Yeast (Saccharomyces cerevisiae) one-hybrid screens have been effective in the isolation of legume proteins that bind specific cis-elements. Sangwan and O'Brian (2002) constructed a soybean nodule cDNA library in a vector containing a GAL4-activation domain to produce GAL4 fusion proteins. Introduction of this library into a yeast His-auxotroph engineered to have a HIS3 gene preceded by a (GA)₂₇/(CT)₂₇ dinucleotide repeat sequence resulted in the identification of a His prototroph containing a GBP-GAL4 fusion. Rarely have legume TF protein-protein interactions been identified, presumably because most of the research in this area has focused on protein-DNA interactions. Kim et al. (2001) identified a C2H2-type Zn-finger protein, SCOF-1, which failed to exhibit DNA-binding activity with several candidate cis-elements, but was subsequently found to enhance the abscisic acid response element-dependent gene expression mediated by SGBF-1.

Approaches to demonstrate DNA binding of legume TFs include electrophoretic mobility shift assays (EMSAs), hybridization of labeled DNA to TFs on filters, DNase-I footprinting, and yeast one-hybrid assays (Table III). In one interesting example, Bastola et al. (1998) identified cDNA encoding Alfin1, a protein with a putative Zn-binding domain, by differential screening of salt-tolerant alfalfa cells. The DNA-binding specificity of Alfin1 was determined by binding of purified protein to random oligonucleotides in an EMSA followed by PCR amplification to identify the preferential target sites. In most instances, however, DNA-binding specificity has been tested only on a few select cis-element sequences that have been identified by work in other plant families. Such biased approaches are likely to miss important TF-DNA interactions. Alternative, nonbiased approaches are now available that should solve this problem (see below).

Further evidence of TF activity has occasionally been provided using transactivation assays. Some groups have demonstrated in vivo transactivation in cell culture and transient transformation systems, including particle bombardment of bean cotyledons (Chern et al., 1996a, 1996b; Bobb et al., 1997), polyethylene glycol-mediated transfection of Arabidopsis protoplasts (Kim et al., 2001), and in vitro transcription activation in rice (Oryza sativa) cell extracts (Lindsay et al., 2002).

One verification of a protein's role as a TF is its localization to the nucleus. For legume TFs, this has been done with immunohistochemical localization (Rodriguez-Uribe and O'Connell, 2006) or using GFP or GUS fusion proteins (Kim et al., 2001; Kaló et al., 2005; Smit et al., 2005; Wang et al., 2005; Shukla et al., 2006). In the case of NSP2, a regulator of legume-rhizobium symbiosis, nuclear relocalization was detected following application of purified Nod factors, suggesting that posttranslational modification is required to activate this protein (Kaló et al., 2005).

So far, there has been a major disconnect between TFs that have been ascribed a biological role based on genetic data and TFs that have been characterized at the biochemical and/or molecular levels. Clearly, to understand better the function of genetically characterized TFs, we need to identify the genes and network of genes that they control. On the other hand, for TFs that have been characterized in terms of their DNA-binding ability, it is now important that biological function be established via forward or reverse genetics. Furthermore, despite the knowledge that TFs often work as part of a team or complex of proteins to recruit or block recruitment of RNA polymerase to the DNA (Lee and Young, 2000), virtually nothing is known about the proteins that interact with legume TFs to ensure their biological activity. Tissue and organ development and differentiation and plant responses to specific environmental challenges require the concerted activity of networks of TFs, which orchestrate global changes in transcription. The details of these networks remain unknown in legumes. Addressing these open questions in legume TF biology in an efficient manner will require a coordinated effort on the part of the scientific community. The final section of this review offers a roadmap for this enterprise.

A ROADMAP FOR FUTURE RESEARCH ON LEGUME TFs

Within the next 2 or 3 years, complete or near-complete genomic sequence for the euchromatic regions of three legumes, Medicago, Lotus, and soybean, will be available. This will greatly facilitate systematic approaches to TF functional analysis. Bioinformatics approaches will rapidly identify putative TFs among the new genomic sequences, as described above, which will provide grist for the functional analysis mill. Whereas forward genetics will gather momentum as genomic sequencing results in more complete and better integrated physical and genetic maps of chromosomes, which will facilitate map-based cloning of additional TFs involved in legume development and differentiation, reverse-genetics approaches are likely to play a more significant role in the functional characterization of TFs in the future. Certainly, reverse genetics offers a more systematic way to characterize all putative TF genes.

Some of the tools for systematic reverse-genetics analysis of TF function in legumes, such as plant transformation protocols for RNAi and overexpression (Thykjaer et al., 1997; Chabaud et al., 2003; Ott et al., 2005; Zhang et al., 2005) and ethyl methanesulfonate mutant populations for TILLING (Perry et al., 2003), are already in place. Others, such as transposon-insertion (Tadege et al., 2005) and fast neutron bombardment deletion mutant populations (Wang et al., 2006) are being developed, and a mutant from a small Tnt1 insertion population has already been described (Benlloch et al., 2006). Viral-induced gene silencing is another promising tool for high-throughput reverse genetics that has proven successful in pea (Constantin et al., 2004).

In view of the TF content of Arabidopsis and rice, we expect that each of the three model legumes mentioned above will possess at least 2,000 TF genes. It will be an impossible task for any one group to characterize this number of genes, at least at the genetic level. A coordinated international effort would help to make the process of TF gene function discovery most efficient. One way to give direction to such an enterprise would be to determine first the developmental and environmental expression profiles of each TF in the context of the whole transcriptome. This would serve several purposes. First, it would reveal any organ/developmental specificity. Second, it would reveal any environmental stress specificity, which would constrain hypotheses about possible roles of each TF. Third, by setting TF gene expression profiles into the broader, whole-genome context of transcription, correlations between individual TFs and groups of other genes would be revealed, which would help to refine hypotheses about possible TF function, especially if correlated sets of genes are predicted to be involved in one or just a few biological processes. Many of the tools required for such transcriptome analyses are now available for Medicago, Lotus, and soybean, including Affymetrix GeneChips containing probe sets for the majority of genes in these three models. In addition, we are currently developing gene-specific primers for all Medicago TFs for qRT-PCR to complement data obtained using the corresponding Affymetrix GeneChip (K. Kakar and M.K. Udvardi, unpublished data), and a similar resource is being developed for soybean (G. Stacey, personal communication). Transcript quantification by qRT-PCR is more sensitive than by DNA array hybridization methods (Czechowski et al., 2004) so the resources being developed for qRT-PCR profiling in Medicago and soybean will provide a more comprehensive picture of TF expression patterns in this species. Hierarchical cluster analysis (Yu et al., 2005) and Pearson correlation (Zar, 1999; Persson et al., 2005) are two ways to identify genes that are coordinately regulated, which will not only provide clues about the possible function of TF genes (see above), but also identify possible downstream target genes of specific TFs.

There have been few attempts to confirm the physical interaction between a genetically characterized legume TF and a target gene, although possible target genes have been identified by transcriptome analysis of TF mutants (e.g. Kaló et al., 2005; Smit et al., 2005). Methods have been developed to identify TF target genes in a nonbiased, high-throughput manner. Perhaps the most powerful of these is chromatin immunoprecipitation (ChIP) followed by DNA array hybridization (called ChIP-chip) to identify DNA fragments covalently bound to immunoprecipitated DNA-binding proteins (Thibaud-Nissen et al., 2006). Specific immunoprecipitation can be facilitated by in planta expression of the TF of interest as a fusion protein with a short, nonplant peptide epitope at one end, which enables the use of commercially available monoclonal antibodies directed toward the epitope to precipitate the fusion protein and any covalently linked proteins and DNA. This approach has been used, for example, to identify genomic DNA bound by the Arabidopsis FLC TF (Helliwell et al., 2006). Affinity purification can also be used to identify associated proteins in DNA-binding complexes (Wood et al., 2006). A prerequisite for ChIP-chip is an array containing probes for promoter DNA. Arrays designed to detect gene transcripts, such as the Affymetrix GeneChips for Medicago, Lotus, and soybean, are not suitable for such applications. However, tiling arrays, which contain oligonucleotide probes covering the entire genome (coding and noncoding) with short or no gaps between probed sequences are well suited to ChIP-chip (Thibaud-Nissen et al., 2006; Zhang et al., 2006). Although no tiling array exists for a legume yet, we have plans to develop a Medicago tiling array upon completion of the genome sequence in 2008.

The preceding paragraphs may give the impression that the road is mostly clear for rapid progress in TF function discovery in model legumes. However, it is likely that there will be bumps, potholes, and unexpected turns in the road ahead. For instance, some TFs appear to job share with one or more close relatives, so that loss of function of one gene may go unnoticed in a mutant plant (Riechmann and Ratcliffe, 2000). Functional redundancy between some TFs makes it difficult, if not impossible, to isolate mutations in these genes via forward genetics. However, reverse-genetics approaches that utilize phylogenetic and transcriptomic information to identify potentially redundant genes prior to the creation of double or higher order mutants should overcome this problem (e.g. Liljegren et al., 2000; Zhang et al., 2003). Obviously, the availability of well-curated mutant populations (see above) will be essential for this endeavor. An alternative approach to overcome functional redundancy is to create dominant-negative mutants by fusing TFs to known repressor domains (Markel et al., 2002). Yet another approach is to overexpress the TF of interest in transgenic plants and to monitor the effects of this on the expression of other genes and on the phenotype (biochemical, physiological, developmental, or otherwise) of the altered plants. Ideally, TF overexpression should be confined to the same cell, tissue, and organ types as the endogenous gene and preferably under the control of an inducible promoter. There is mounting evidence that diversification of TF function in plants often results from changes in nontranscribed sequences that alter the expression domain of the gene, rather than from changes in the coding sequence that alter the DNA- or protein-binding properties of TFs. Thus, TF mutant phenotypes have been suppressed by ectopic expression of related TFs that are not normally expressed in the same tissue/organ as the mutated TF (Riechmann, 2002). Overexpression of TFs can also interfere with processes totally unrelated to the normal function of the protein (Riechmann and Ratcliffe, 2000). Clearly, care must be taken in designing and interpreting TF overexpression experiments.

TFs interact physically with other proteins, in addition to the RNA polymerase complex itself, to effect changes in gene transcription (Lee and Young, 2000). The specific makeup of these complexes is unknown for the majority of plant genes, although this knowledge is a prerequisite to understanding the combinatorial control of transcription (Singh, 1998). Approaches such as yeast two-hybrid screening (e.g. Zhang et al., 1999) and ChIP followed by proteomic analysis of the resulting protein complexes (Helliwell et al., 2006; Wood et al., 2006) will be useful in this context. Finally, the network of genes regulated by a single TF and its partners is just a small part of a larger genetic regulatory network that ensures coordinated expression of genes involved in many different cellular processes during plant development and differentiation. Deciphering these global genetic networks will require integration of many of the genomic, functional genomic, and bioinformatic approaches described above.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Table S1. Putative TFs among IMGAG-annotated proteins.