Inactivation of the Arabidopsis AGL5 MADS-box gene by homologous recombination

Sherry A. Kempin1, Sarah J. Liljegren1, Laura M. Block1, Steven D. Rounsley1, Eric Lam2, and Martin F. Yanofsky1

1 Department of Biology and Center for Molecular Genetics, University of California at San Diego, La Jolla, CA 92093-0116, USA

2 AgBiotech Center, Foran Hall, Dudley Rd., Rutgers the State University of New Jersey, New Brunswick, NJ 08903, USA

Supported by the National Science Foundation

MADS-box genes encode putative transcription factors that have been shown to play important regulatory roles in plants, yeast and animals1. In plants, MADS-box genes regulate key events in flower development, such as the specification of flower-meristem and flower-organ identity, as well as the differentiation of each flower organ2,3. In Arabidopsis, more than two dozen MADS-box genes have been cloned, and yet the functional analyses of many of these genes awaits the isolation of mutant alleles. These studies have also been impeded in part by the redundancy exhibited by closely related members of the MADS-box gene family4,5. For example, the APETALA1 (AP1) and CAULIFLOWER (CAL) MADS-box genes have overlapping functions that specify flower-meristem identity. cal single mutant plants are indistinguishable from wild-type plants, however, mutations in the CAL gene dramatically enhance the phenotype of ap1 single mutants6-8. The functional redundancy observed between the CAL and AP1 genes has been attributed to their closely related sequences and expression patterns, suggesting that they regulate many of the same targets.

Many other plant MADS-box genes in addition to AP1 and CAL likely exhibit functionally redundant activities. The carpel- and ovule-specific AGL1 and AGL5 genes are also closely related at the sequence level, and their expression patterns are remarkably similar9-11. Because AGL1 and AGL5 may have overlapping activities, it is likely that mutations in these genes may have been missed in standard mutant screens, simply because the single mutants may not reveal a significantly altered phenotype. As a start toward circumventing this problem, we have targeted the AGL5 gene for inactivation by homologous recombination. We have previously shown that AGL5 represents a candidate target gene regulated by the flower-organ-identity MADS-box gene AGAMOUS (AG)10,12,13. The onset of AGL5 expression is shortly after that of AG within carpel primordia, and the normal expression pattern of AGL5 requires the prior activity of AG. Furthermore, AG binds to a target sequence within the AGL5 promoter and ectopic activity of AG results in ectopic expression of AGL510. These and other data suggest that AGL5, perhaps in combination with AGL1, functions downstream of AG to program normal carpel and ovule development.

Experiments in Arabidopsis tissue-cultured cells using a root transformation procedure suggested that targeted disruption of cloned genes is possible14. However, the inability of this cell line to be regenerated prohibited the isolation of a homozygous mutant plant, and thus subsequent analyses were not possible. Recent improvements in intact plant transformation technology15 have provided a more efficient method for generating large numbers of transgenic plants, suggesting that it should be possible to identify a targeted insertion, even if the frequency of such an event is relatively low.

To test this approach, and to begin to investigate the functional role of the AGL5 MADS-box gene in Arabidopsis, we have used a PCR-based assay of transgenic plants to identify targeted insertions into AGL5. The targeting construct consisted of a kanamycin-resistance cassette that was inserted between approximately 3 kb and 2 kb segments representing the 5' and 3' regions of the AGL5 gene, respectively (Fig. 1). If a successfully targeted insertion were to occur by homologous recombination, the result would be a 1.6 kb deletion within the AGL5 gene. Specifically, the targeted allele could potentially encode only the first 42 of 246 amino acid residues, and should include only 26 of the 56 amino acids comprising the DNA-binding MADS-domain. In addition, the recombination event would result in the insertion of the 2.5 kb kanamycin-resistance cassette within the AGL5 coding sequence.

750 kanamycin-resistant transgenic lines were produced by Agrobacterium-mediated transformation, and pools of transformants were analyzed using a PCR assay (see methods) to determine if any of these primary transformants had generated the desired targeted insertion into AGL5. A single line was identified that appeared to contain the anticipated insertion, and this line was allowed to self-pollinate to permit further analyses in subsequent generations. Genomic DNA from the homozygous mutant plants was analyzed with two different restriction enzymes (Fig. 3) and by several distinct PCR amplifications (Fig. 2), and all data were consistent with the desired targeting event. We also analyzed the regions flanking the AGL5 gene to verify that there were no detectable deletions or rearrangements of sequences outside of AGL5 (Fig. 4 and Fig. 5). Because the kanamycin-resistance cassette within the AGL5 targeting construct contains sequences that specify transcription termination14, we anticipated that little or no AGL5 RNA would be detected in the homozygous mutant plants. Using a probe specific for the 3' portion of the AGL5 cDNA, we detected AGL5 transcripts in wild-type but not in agl5 mutant plants (Fig. 6). These data suggest that, as anticipated, the targeted disruption of the AGL5 gene likely represents a complete loss-of-function allele.


Targeted disruption by homologous recombination provides a new tool along with other methods, such as T-DNA insertion and transposon tagging, for obtaining loss-of-function (knockout) alleles of cloned genes in Arabidopsis. Homologous recombination has a number of advantages over these other methods since this procedure can be relatively rapid (approximately six weeks from the time Arabidopsis plants are transformed to the time that a targeted disruption can be identified) and allows precise construction of both knockout and more subtle "knockin" mutant alleles.

One of the important questions yet to be answered is the frequency with which homologous recombination can be used to generate specific mutations. Our results here with AGL5, together with the previous study14 suggest that the frequency may be in the range of 10-3. However, additional studies will be needed to make an accurate estimate and to determine if the frequency varies for different regions of the genome. Because our targeting construct contained relatively small DNA fragments (2 kb and 3 kb), this demonstrates that large DNA fragments are not required. However, based on targeting efficiencies in mammalian systems, it seems likely that the overall efficiency will be improved by using larger DNA segments.

One of the major unanswered questions is the mechanism by which Agrobacterium-mediated transformation, which normally integrates its transferred DNA (T-DNA) using a non-homologous mechanism, can lead to DNA integration by homologous recombination. The proposed mechanism for T-DNA transfer involves the transfer of a single stranded intermediate (the T-strand), which is coated by the VirE2 single stranded DNA-binding protein, and has the VirD2 endonuclease covalently attached to its 5' end20. It is formally possible that this DNA/protein complex serves as the substrate for the relatively rare homologous events, in addition to the typical non-homologous T-DNA integrations. However, another possibility is that the T-DNA first must integrate at a nearby site (in our case, near AGL5), and that this initial non-homologous integration then serves as a substrate for the subsequent homologous recombination event.

Although in the long run, it may prove useful to include negative-selection strategies to improve the ratio of homologous to non-homologous events, in the near future it might be wise to do so with caution. For example, if the mechanism involves the initial integration at a nearby site followed by the homologous recombination event, the primary transformant is likely to contain both homologous and non-homologous integrations. Furthermore, many primary transformants contain T-DNA insertions at multiple sites. Thus, use of a negative-selection strategy of primary transformants could result in the loss of desirable events, which otherwise could easily be segregated away from non-homologous integration events.


Special thanks go to Doug Geerdes and Tina Kuhlman for technical assistance.

| References | Methods | Home |