RNAi for Plant Functional Genomics

A major challenge in the post-genome era of plant biology is to determine the functions of all the genes in the plant genome. A straightforward approach to this problem is to reduce or knock out expression of a gene with the hope of seeing a phenotype that is suggestive of its function. Insertional mutagenesis is a useful tool for this type of study, but it is limited by gene redundancy, lethal knock-outs, nontagged mutants and the inability to target the inserted element to a specific gene. RNA interference (RNAi) of plant genes, using constructs encoding self-complementary ‘hairpin’ RNA, largely overcomes these problems. RNAi has been used very effectively in Caenorhabditis elegans functional genomics, and resources are currently being developed for the application of RNAi to high-throughput plant functional genomics.


Introduction
Plant genome and EST sequencing efforts have yielded an abundance of genetic information, including the complete genomic sequences of Arabidopsis thaliana and rice. Genome sequencing projects are also currently under way for maize, Medicago truncatula and Lotus japonicus, among others. In addition, a large number of EST sequencing projects are being pursued in a variety of plants. This wealth of information has created a need for efficient, high-throughput methods to utilize such data for functional genomic analyses. This need is well illustrated by the fact that less than 10% of the predicted 28 581 genes in the Arabidopsis genome [39] have been experimentally characterized, and 30% lack significant homology to characterized genes from any species [2].
In plants, RNA interference [RNAi, also known as post-transcriptional gene silencing (PTGS) or co-suppression] is thought to be a key defence against viruses (reviewed in [44]), as well as a way of regulating endogenous genes (reviewed in RNAi for plant functional genomics 241 recently the knock-down phenotypes of approximately 86% of the predicted C. elegans genes were analysed using this approach [23]. From these studies, it is clear that RNAi is a precise method for investigating gene functions in C. elegans. The primary advantage of RNAi in these studies is the simplicity of the techniques involved -feeding C. elegans with E. coli expressing dsRNA [41] or simply soaking worms in dsRNA [38] will knock down the homologous C. elegans gene. Results from large-scale RNAi investigations appear to integrate well with other large-scale approaches, specifically transcription profiling with microarrays and yeast two-hybrid studies of protein-protein interactions [15,25]. From these coordinated efforts, comprehensive pictures of the C. elegans genes involved in important biochemical and morphogenetic pathways are emerging.

Mutagenesis-based functional genomics in plants
In contrast to the situation in C. elegans, highthroughput expression modulation techniques are not so easily applicable to plants. Current techniques of altering gene expression and function include chemical mutagenesis using ethylmethane sulphonate (EMS), insertional mutagenesis using transposons or T-DNAs and RNAi (reviewed in [4]). The more recently developed techniques utilizing RNAi will be discussed below. Chemical and insertional mutagenesis approaches have been applied extensively to the functional characterization of plant genes, and many successful examples of their uses have been published. However, these approaches have disadvantages stemming from the randomness of mutagenesis. Mutations must be sequenced or mapped to confirm their positions, and large collections of mutant lines are necessary to obtain good coverage of whole genomes. These limitations mean that the application of mutagenesis on a genomic scale is very labourintensive.
Insertional mutagenesis, rather than chemical mutagenesis, is more commonly used in largescale plant functional genomics, as mobile DNA tags of known sequence can be utilized to retrieve flanking gene sequences. Several databases of T-DNA and transposon insertion lines have been compiled, e.g. the Syngenta Arabidopsis Insertion Library (SAIL), which contains T-DNA flanking sequences from 52 964 lines [34], and the Salk Institute Genomic Analysis Laboratory (SIGnAL) collection of T-DNA insertions [33], which contains 88 122 T-DNA insertions, covering approximately 74% of the predicted Arabidopsis genes [1]. The large T-DNA insertion collections required to achieve moderate coverage of Arabidopsis genes reveal the primary limitation of random mutagenesis -that it cannot be targeted to specific plant DNA sequences. The possibility of multiple transposon or T-DNA insertions confounding the analysis of mutant lines is a further problem of insertional mutagenesis, e.g. lines in the SIGnAL collection have an average of about 1.5 T-DNA insertions/line [1] and in the SAIL collection an average of 1.5-2 T-DNA insertions/line [37]. Also, the fact that recessive mutations will only manifest their phenotypes in segregating M2 populations further complicates large-scale analysis of insertion lines.
The limitations arising from the randomness of mutagenesis are exemplified by the current coordinated effort to characterize all of the R2R3-MYB gene family members using insertional mutants [26,30,36]. There are 125 members of the R2R3 sub-family of MYB transcription factors in Arabidopsis [36]. By screening for T-DNA and transposon insertions in 73 of these genes, 47 insertions in 36 different target genes were identified [30]. This involved laborious screening -over 50 000 mutagenized lines were screened for insertions in subsets of the MYB genes. Therefore, while insertional mutagenesis has been shown to be very effective in terms of revealing mutant phenotypes, its applicability to high-throughput reverse-genetic analysis of plants is limited.

RNAi for functional genomics in plants
In contrast to insertional mutagenesis, the interruption of gene expression by RNAi has only been utilized for plant functional genomics relatively recently. As discussed above, dsRNA triggers the degradation of homologous RNAs, such as mRNAs, thus preventing gene expression. dsRNAs are efficiently produced by intron-spliced hairpin transgenes (sense and antisense copies of the target gene separated by an intron), which have been found to be extremely effective in triggering RNAi [35]. dsRNAs can be delivered to plants in several ways (reviewed in [43]): microprojectile bombardment with dsRNA or intron-containing hairpin RNA (ihpRNA)-expressing vectors; infiltration of plant tissue with an Agrobacterium strain carrying a T-DNA expressing an ihpRNA transgene; virusinduced gene silencing (VIGS), in which the target sequence is integrated into viral sequences which are used to infect the plant, or are expressed from Agrobacterium-introduced transgenes, and by stable transformation with ihpRNA expressing transgenes. The various RNAi techniques each have advantages and disadvantages with respect to how persistent their effect is and the range of plants to which they can be applied, e.g. bombardment can be applied to any plant, but produces only transient effects. Alternatively, transformation with ihpRNA-expressing transgenes provides stable and heritable gene silencing, but requires efficient plant transformation techniques. ihpRNA transgenes have been shown to be very effective for a wide range of target genes in various plant species (reviewed in [43,45]), indicating that the RNAi mechanism is probably conserved in all plant species. This is supported by a recent report of RNAi in the non-vascular moss Physcomitrella patens [5].
RNAi has several advantages over insertional mutagenesis strategies for functional genomics, the primary advantage being an ability to specifically target the chosen gene. As RNAi is a homologydependent process: careful selection of a unique region of the target sequence can ensure that a specific gene family member is silenced, or multiple members of a gene family can be silenced by targeting highly conserved sequence domains. In this way, redundancy is not limiting. RNAi can also be used to analyse the functions of essential genes. Variable levels of gene silencing can be achieved in different transgenic lines using the same ihpRNA construct [45], allowing selection of lines with a greater or lesser degree of silencing. Also, the expression of ihpRNAs from inducible promoters can control the extent and timing of gene silencing [7,16], such that essential genes are only silenced at chosen growth stages or in chosen plant organs. In these ways, RNAi provides the flexibility necessary for the characterization of genes of diverse functions.

RNAi tools for functional genomics
Comprehensive resources for functional genomics in plants are being developed within the research community, the most notable examples being the sequencing of the Arabidopsis and rice genomes [2,13,46]. Utilization of the Arabidopsis genome is being made possible through efforts such as that of the complete Arabidopsis transcriptome microarray (CATMA) project [10], which aims to make a high-specificity gene sequence tag (GST) library covering all Arabidopsis genes. Compatible tools are also becoming available, such as Curtis and Grossniklaus' suite of binary vectors suitable for constitutive or inducible ectopic expression and expression of GUS or GFP fusion proteins [11], the pHELLSGATE high-throughput gene silencing vectors [19,20] and a high-throughput tobacco rattle virus-based VIGS vector [27]. An increasingly common feature, shared by all of these vectors, is the use of Invitrogen's Gateway recombinationbased cloning technology [18] to replace previously time-limiting conventional cloning steps. The compatibility arising from common use of the Gateway system is currently being exploited by the Arabidopsis genome RNAi knock-out line analysis (AGRIKOLA) consortium to create a pHELLSGATE-based silencing vector for each GST in the CATMA collection, with the aim of producing ihpRNA constructs for all Arabidopsis genes [21]. These vectors will provide an invaluable resource for Arabidopsis functional genomics.

Challenges for the use of RNAi for functional genomics in plants
There are limitations to the ways in which RNAi can currently be used for high-throughput functional genomics across different plant species. First, unlike insertional mutagenesis, knowledge of the target gene sequence is required. However, once sequence information is available, high-throughput approaches can be quickly applied, as is evident with the resources currently being developed for Arabidopsis functional genomics. With the increasing number of genome and EST sequencing projects, sequence knowledge is becoming much less of a limitation. Second, the methods of delivering dsRNA limit the species to which highthroughput approaches can be applied, e.g. transformation is necessary for stable expression of ihpRNA transgenes, and Arabidopsis remains the only plant that can be transformed with the ease necessary for high-throughput approaches [9]. Similarly, VIGS, the main alternative to stable transformation with ihpRNA transgenes, still requires viral vectors compatible with the plant species of choice. To this end, improvements in plant transformation techniques and the development of further VIGS vectors is necessary.
The difficulty of detecting subtle or conditional mutant phenotypes has been reported for largescale C. elegans reverse-genetic screens [12,23]. As could be expected in such studies, the efficiency of detection of mutant phenotypes depends upon how closely, and under what conditions, mutants are examined. Thus, a report of no phenotype is far from confirmation of a gene being redundant or non-functional. For plant screens, several approaches are being used to address this problem, including the development of standards of normal growth and development across a range of Arabidopsis growth stages [6] and the development of sets of growth conditions (with varied environmental stresses) to test for conditional mutants [3]. Inclusion of marker genes in ihpRNA constructs to indicate the strength of gene silencing presents a novel way of screening for subtle phenotypes using RNAi. Such an approach is based upon the discovery that inclusion of sequences from two genes within the one hairpin transgene results in their coordinate silencing [20]. By including a readilyvisualized phenotypic marker, such a system could be used to identify strongly silenced lines which do not display obvious phenotypes. These can then be examined in more detail under varied environmental conditions.

Conclusions
With the comprehensive resources currently being developed, RNAi-based functional genomics is quickly becoming a reality, particularly in Arabidopsis. The utility of RNAi strategies for functional genomics has been demonstrated by one of the first large-scale screens published, which utilized VIGS to investigate the function of almost 5000 random Nicotiana benthamiana cDNAs in Pto-mediated disease resistance [28]. However, more random and extensive screens are clearly necessary to investigate the functions of unclassified or uncharacterized genes. Challenging goals are being set for plant functional genomics by projects such as REGIA [31], which aims to characterize all Arabidopsis transcription factors, and the 2010 Project [8], which aims to functionally characterize all Arabidopsis genes by the year 2010. RNAi clearly provides a significant tool for the accomplishment of these goals, and will undoubtedly be used to address many other challenges in plant functional genomics.