The FOX hunting system: an alternative gain-of-function gene hunting technique


  • Takanari Ichikawa,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Miki Nakazawa,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Mika Kawashima,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Haruko Iizumi,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Hirofumi Kuroda,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Youichi Kondou,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Yumi Tsuhara,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Kumiko Suzuki,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Akie Ishikawa,

    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Motoaki Seki,

    1. Plant Genomic Network Research Team, Plant Functional Genomics Research Group, RIKEN Plant Science Center, RIKEN Yokohama Institute, 1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Miki Fujita,

    1. CREST, JST, Laboratory of Plant Molecular Biology, RIKEN Tsukuba Institute, 3-1-1 Koyadai, Tsukuba 305-0074, Japan,
    Search for more papers by this author
  • Reiko Motohashi,

    1. Faculty of Agriculture, Shizuoka University, 836 Ohya Suruga-ku, Shizuoka 422-8529, Japan, and
    Search for more papers by this author
  • Noriko Nagata,

    1. Faculty of Science, Japan Women's University, 2-8-1, Mejirodai, Bunkyo-ku, Tokyo 112-8681, Japan
    Search for more papers by this author
  • Tomoko Takagi,

    1. Faculty of Science, Japan Women's University, 2-8-1, Mejirodai, Bunkyo-ku, Tokyo 112-8681, Japan
    Search for more papers by this author
  • Kazuo Shinozaki,

    1. Plant Genomic Network Research Team, Plant Functional Genomics Research Group, RIKEN Plant Science Center, RIKEN Yokohama Institute, 1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama, Kanagawa, 230-0045, Japan,
    Search for more papers by this author
  • Minami Matsui

    Corresponding author
    1. Plant Functional Genomics Research Team, Plant Functional Genomics Research Group, Plant Science Center RIKEN Yokohama Institute, 1-7-22 Suehirocho, Tsurumiku, Yokohama, Kanagawa, 230-0045, Japan,
      *(fax +81 45 503 9584; e-mail
    Search for more papers by this author

*(fax +81 45 503 9584; e-mail


We have developed a novel gain-of-function system that we have named the FOX hunting system (Full-length cDNA Over-eXpressing gene hunting system). We used normalized full-length cDNA and introduced each cDNA into Arabidopsis by in planta transformation. About 10 000 independent full-length Arabidopsis cDNAs were expressed independently under the CaMV 35S promoter in Arabidopsis. Each transgenic Arabidopsis contained on average 2.6 cDNA clones and was monitored under various categories such as morphological changes, fertility and leaf color. We found 1487 possible morphological mutants from 15 547 transformants. When 115 pale green T1 mutants were analyzed, 59 lines represented the mutant phenotypes in more than 50% of the T2 progeny. Characterization of two leaf color mutants revealed the significance of this approach. We also document mutants from several categories and their corresponding full-length cDNAs.


Gain-of-function mutants reveal gene functions that would never be uncovered by conventional loss-of-function approaches. In Arabidopsis an activation tagging technique was developed that involves introducing transcriptional enhancers into the plant genome (Jeong et al., 2002; Marsch-Martinez et al., 2002; Walden et al., 1994; Weigel et al., 2000; Wilson et al., 1996). These enhancers activate genes that are located proximal to the insertion sites. An enhancer can affect transcription on both sides of the insertion sites (Ichikawa et al., 2003; Jeong et al., 2006). An advantage of this technology is that a different spectrum of mutants is obtained compared with that attained from conventional knockout methods. We have generated around 55 000 activation-tagged Arabidopsis mutants as a genome research resource (Ichikawa et al., 2003; Nakazawa et al., 2003). Among these mutants we have characterized several dominant mutants that are caused by mutations in genes that are members of a gene family. In plants with a knockout mutation of these genes there are only marginal differences from wild type as a result of gene redundancy (Nakazawa et al., 2001; Zhao et al., 2004). This means it would not have been possible to isolate these activation-tagged mutants by conventional T-DNA or transposon insertion mutagenesis. However, activation-tagged mutagenesis does have some disadvantages. One is that the effect of gene activation by a transcriptional enhancer is not restricted to one gene and in some cases transcription of several genes is increased, resulting in a complex phenotype (Ichikawa et al., 2003). To circumvent this problem, and for systematic analysis of gain-of-function mutations, we generated alternative gain-of-function mutant lines caused by ectopic expression of full-length cDNAs. We introduced around 10 000 independent Arabidopsis full-length cDNAs into an expression vector under the control of the cauliflower mosaic virus 35S (CaMV 35S) promoter. This cDNA expression library was introduced into Arabidopsis by in planta transformation. We have generated about 15 000 transgenic lines that express Arabidopsis full-length cDNAs (Arabidopsis Full-length cDNA Over-eXpressing transgenic line; or Arabidopsis FOX line) and on average 1.2 full-length cDNAs were amplified from each line. Ectopic expression of full-length cDNAs causes dominant mutant phenotypes. While generating these FOX lines we monitored mutant phenotypes under various categories. We present here two leaf color mutants and also several morphological mutants with their corresponding genes. We also discuss this systematic gene activation technique, the Full-length cDNA Over-eXpressor gene hunting system (FOX hunting system) in comparison with the activation tagging method.

Results and discussion

Construction of an Agrobacterium expression library that contains approximately 10 000 normalized Arabidopsis full-length cDNAs

We mixed around 10 000 independent Arabidopsis full-length cDNAs in nearly equal molar ratio to generate a normalized Arabidopsis full-length cDNA mixture. These cDNAs originated from the RIKEN Arabidopsis full-length cDNA collection and each cDNA was sequenced (Seki et al., 2002). As internal standards we also mixed four bacterial oncogenes that are known to induce morphological phenotypes and sterility. These internal controls can be used as an indication of cDNA representation in the library but after transformation into Arabidopsis these oncogenes can be eliminated from the mutant lines in their second generation. We used tms1 (Klee et al., 1984), iaaM (Comai and Kosuge, 1982), rolB (Furner et al., 1986) and tmr (Lichtenstein et al., 1984). They are known to have drastic effects on the morphology of many plants (Casanova et al., 2005; Romano et al., 1995; Smart et al., 1991). These four oncogenes were added to the cDNA mixture at nearly double the molar ratio of each full-length cDNA. It was expected that one of these internal controls would be encountered in every 7500 clones in the population.

After the Agrobacterium library (FOX Agrobacterium library) was made, T-DNA-specific primers were designed to perform PCR on DNAs isolated from randomly chosen Agrobacterium colonies. When PCR fragments from 34 Agrobacterium colonies were amplified, 20 colonies contained a single cDNA construct, 11 contained two different cDNA constructs, one contained three different cDNA constructs and two colonies contained the empty vector (Figure 1b). The average size of these 45 amplified cDNA fragments was 1.4 kb in the range 0.3–3.0 kb. When the 45 fragments were sequenced, all cDNAs were found to be independent of each other (data not shown).

Figure 1.

 Distribution of full-length cDNA transgenes in agrobacteria and in plants.
(a) Construct used for full-length cDNA expression in Arabidopsis. RIKEN Arabidopsis full-length (RAFL) cDNAs from lambda libraries were recloned into pBIG2113SF, a plant binary vector. El, 5′-upstream sequence of CaMV 35S promoter (−419 to −90); P35S, CaMV 35S promoter (−90 to −1); Ω, 5′-upstream sequence of TMV; NOS-T, polyadenylation signal of the gene for nopaline synthase in the Ti plasmid; Hyg, hygromycin resistance gene. Arrows indicate the positions of the GS4 and GS6 primers used to recover the cDNAs.
(b) Example of size distribution of the RAFL cDNAs in agrobacteria transformed with the full-length cDNA-expressing T-DNA library. After a first electrophoresis RAFL cDNA fragments were sorted according to their lengths. Lane M, Lambda/HindIII markers. In lanes 1, 2, 5, 6, 7, 9, 10, 13, 15, 17, 19 and 22 several bands were amplified.
(c) Size distribution of the 106 RAFL cDNA fragments amplified from the Arabidopsis FOX plants (black bars) compared with 277 randomly chosen RAFL cDNAs in the Lambda vector (white bars). In order to emphasize the similarity of the distribution pattern values from the Lambda vector were divided by two before plotting.

Generation of Arabidopsis FOX lines

The FOX Agrobacterium library was used to transform the Arabidopsis Columbia-0 (Col-0) ecotype to generate Arabidopsis FOX lines. After floral dipping, seeds from the T0 plants were collected and they were selected on basic agar medium (BAM) plates containing hygromycin (Nakazawa and Matsui, 2003). These plates select transgenic plants within a week of germination and there is no waiting for further incubation to select hygromycin-resistant seedlings. The hygromycin-resistant T1 seedlings were transplanted to soil and over 15 000 Arabidopsis fertile FOX lines were generated. Theoretically, random display of 15 000 lines from the population of 10 000 independent clones should cover 77% of the total population. To confirm the presence of the transgenes, 24 randomly chosen plants were analyzed by the Southern method for the presence of the marker gene using the hygromycin gene as a probe. All lines contained the hygromycin resistance gene and we found on average 2.6 T-DNA inserts in these Arabidopsis FOX lines (data not shown).

Size distribution and sequence difference of full-length cDNAs in Arabidopsis FOX lines

Since full-length cDNAs were integrated into the expression vector and the size distribution was 0.3–3 kb with an average of 1.4 kb it is important to know whether this distribution was reflected in the Arabidopsis FOX lines. In order to estimate the size distribution and variation of the integrated cDNAs in these lines PCR was performed on genomic DNA isolated from randomly chosen FOX plants using T-DNA-specific primers. The size distribution in 106 lines was 1.4 kb on average within the range 0.3–4.2 kb (Figure 1c). The average number of PCR fragments amplified from these lines was 1.2 per plant. We examined DNA gel-blot analyses and found there were on average 2.6 T-DNA insertions per line. However, only an average of 1.2 PCR fragments were rescued from FOX plants. This lower PCR fragment recovery can be partially explained by the existence of the empty vector in the population, but mainly by duplicated T-DNA integration events such as T-DNA tandem or inverted repeats (De Buck et al., 1999; De Neve et al., 1997; Krizkova and Hrouda, 1998).

The average size and range of cDNA distribution was quite similar to that observed in the FOX Agrobacterium library and also similar to that observed in 277 randomly chosen Arabidopsis full-length cDNAs (Figure 1c).

Forty PCR fragments were sequenced and they were derived from different full-length cDNAs and none of these cDNAs was identical to the bacterial oncogenes added to the cDNA mixture as controls (Table S1).

Monitoring phenotypes in the Arabidopsis FOX lines

In the course of generating 15 547 T1 FOX lines any visible phenotypes were monitored. These phenotypes included alterations in visible morphology, growth rate, plant color, flowering time and fertility. We collected 1487 phenotypically altered lines. We compared the appearance of any morphological mutants with those that appeared in Arabidopsis activation-tagged lines. The distribution of mutants in the various categories was almost the same as in the activation-tagged mutants, but the appearance of mutations was much higher (9% of total plants transplanted) compared with that of the activation tagging population (2% of total plants transplanted) (Ichikawa et al., 2003; Nakazawa et al., 2003). This indicated the higher efficiency of the FOX lines and is mainly due to proper expression of the full-length cDNAs under the control of the strong CaMV promoter and nopaline synthetase (NOS) terminator. To confirm the heritability of the mutant phenotypes in the T2 generation we grew 115 pale green mutant lines that appeared in the T1 generation and looked for the mutant phenotype in the T2 generation (Table S2). We grew 20 seedlings of each to monitor their phenotype. Although it is difficult to precisely establish dominancy from 20 seedlings, 59 lines showed a pale green phenotype in dominant or semi-dominant fashion. We also examined 40 morphological mutants and found only seven lines that represented the original mutant phenotype. However, the percentage and degree of the phenotypic appearance in the T2 generation varied. Most of them appeared in a dominant or semi-dominant fashion, but in some lines the phenotype appeared in only a small percentage of the progeny. This may due to suppression of the transgene in the T2 generation or mischaracterization of the mutant phenotype in the T1 FOX plants. In particular, dwarf and leaf shape mutants could be caused by environmental stress after transfer from selection plates.

Morphology caused by oncogenes

Four oncogenes, tms1, iaaM, rolB and tmr, were mixed in the full-length cDNA library as internal controls. iaaM and tms1 code for tryptophan 2-monooxygenase from Pseudomonas syringae and Agrobacterium tumefaciens, respectively, and over-expression of these gene causes an enhanced auxin response (Comai and Kosuge, 1982; Klee et al., 1984). rolB is from Agrobacterium rhyzogenesis and is involved in sensitivity to auxin. tmr is an oncogene from A. tumefaciens and functions in the biosynthesis of cytokinin. We mixed each of these oncogenes at a ratio of two molar equivalents compared to the other full-length cDNAs.

To monitor the representation and distribution of full-length cDNAs we examined the appearance of these oncogenes in the Arabidopsis FOX lines. First we monitored the morphology of the transgenic plants for mutants caused by these oncogenes. Only two give a morphological phenotype in Arabidopsis. Since both tms1 and iaaM cause an enhanced auxin response, plants containing these genes have accelerated apical dominance and a compact phenotype (Figure 2). These morphological differences are very apparent and easily distinguished from other morphological mutations. Over-expression of tmr causes lethality to plant seedlings and over-expression of rolB causes no obvious phenotype. There were 13 tmr1 or iaaM over-expressing mutants out of 15 547 Arabidopsis FOX lines. The appearance of these mutants was higher than the theoretical value (4 out of 15 000). This may be due to the growth difference of transformed agrobacteria during bacteria cultivation. One of the advantages of these oncogenes, except for rolB, is that they are not carried over to next generation because the mutants are sterile or lethal.

Figure 2.

 Over-expression phenotypes of oncogenes in Arabidopsis.
(a) Phenotype comparison of an iaaM over-expressing plant and a wild-type plant grown for the same period.
(b) Close image of the iaaM over-expressing plant.
(c) tms1 over-expressing phenotype.

Characterization of two leaf color mutants to evaluate the FOX hunting system

One of the major differences between activation-tagged lines and Arabidopsis FOX lines is the expression profiles of the genes. In the FOX lines each full-length cDNA was expressed ectopically by the CaMV 35S promoter. However, activation tagging can represent the original tissue specificity of the enhanced gene expression (Jeong et al., 2002; Weigel et al., 2000). Ectopic expression often causes mis-expression of transgenes and the phenotype that appears with ectopic expression is not necessarily related to the original function of the transgene (Kandasamy et al., 2002; Newman et al., 2004).

Of the 59 lines that showed the pale green phenotype in the T2 generation two lines were chosen as representatives. F03024 plants showed a pale green and late growth phenotype (Figure 3a) and the pale green phenotype was semi-dominant in the T2 generation after self-pollination (Figure 3b,c). F01907 was also isolated as a pale green mutant and its phenotype was dominant in the T2 generation (Figure 3e).

Figure 3.

 Phenotypes related to the F03024 and F01907 lines.
(a) The phenotype of F03024 in the T1 generation.
(b) A plant from the T2 generation of F03024 showing the partial pale green phenotype.
(c) The adult stage of the F03024 line. Left: a F03024 plant showing pale green stems in the T2 generation. Right: a wild-type segregant of F03024 in the T2 generation.
(d) Left: an untransformed wild-type plant. Right: a wild-type plant transformed with the AtPDH1 gene showing the pale green phenotype grown for the same period.
(e) The phenotype of F01907 in the T1 generation.
(f) At3g55240 transformants. Left: an untransformed wild-type plant. Right: an At3g55240 transformant in the T2 generation showing the pale green and tall height phenotype grown for the same period.
(g) Leaf chlorophyll content of: 1, wild-type plants; 2, T1 plants of an At5g55240 transformant; 3, T1 plants of an AtPDH1 transformant. Y-axis: relative chlorophyll content. X-axis: individual plant types.
(h) Average photosynthetic activity of leaves from AtPDH1 and At3g55240 transformants. Lanes 1–4, four independent transgenic plants of AtPDH1; lanes 5–7, three independent transgenic plants of At5g55240. Y-axis: relative photosynthetic activities represented in relation to wild type. X-axis: individual plants.

We recovered the introduced full-length cDNAs by genomic PCR using a T-DNA primer set. From F03024 and F01907 plants we recovered 3.0-kb and 0.8-kb cDNA fragments, respectively. The primers were designed to amplify inserted full-length cDNA fragments. These recovered cDNA fragments were cloned back into the Agrobacterium Ti plasmid vector and the reconstructed plasmids were used to transform Arabidopsis (Col-0) plants. After growth in soil, of 48 transformants generated using the cDNA recovered from F03024, 32 showed the pale green and retarded growth phenotype seen in the original F03024 plants (Figure 3d).

Forty-seven of 51 transformants containing the cDNA recovered from F01907 showed the pale green and early flowering phenotype under long daylight conditions, recapitulating the original F01907 phenotype (Figure 3f).

We sequenced the recovered Arabidopsis full-length cDNAs from F03024. The cDNA from F03024 was At1g70070 (RIKEN Arabidopsis full-length cDNA number AF387007) and encodes a DEVH box helicase (Figures S1a,b and S2a). At1g70070 was over-expressed in F03024 (Figure 4a). DEVH box helicases are members of a gene family that include the DEAD and DEAH box RNA helicases. We named the cDNA recovered from F03024 as AtPDH1 (Arabidopsis prokaryotic DEVH box helicase 1). There is a report that the tobacco DEAD box helicase (VDL) targets the chloroplast and regulates chloroplast development (Wang et al., 2000). AtPDH1 has a putative chloroplast target signal in its N-terminal region belonging to a small clade composed of prokaryotic DEAD box helicases (Figure S2a). We examined the intracellular localization of this protein by fusing the N-terminal 98 amino acids to the green fluorescent protein (GFP) and found it localized in chloroplasts (Figure S3a,b). When photosynthetic activity and chloroplast structure from the pale green leaves of AtPDH1 over-expressing plants were examined, the photosynthetic activity was significantly reduced (Figure 3h); this is due to insufficient development of the chloroplast inner membrane structure (Figure S3c).

Figure 4.

 Detection of levels of AtPDH1 and PEL transcripts by semi-quantitative RT-PCR.
(a) Upper bands are the PCR fragments specific for AtPDH1. Lower bands are the beta-tubulin PCR fragments used for loading adjustment. Lanes 1 and 2, wild-type Columbia plants; lanes 3 and 4, F03024 plants in the T2 generation showing the pale green phenotype.
(b) Reverse transcriptase-PCRs. Upper bands are the PCR fragments specific for At3g55240. Lower bands are the AHA1 PCR fragments used for loading adjustment. Lane 1, wild-type Columbia plant; lanes 2–7, T2-R01907 plants showing the pale green phenotype; lane 8, T2-F01907 plant showing the pale green phenotype.
(c) Reverse transcriptase-PCRs from wild-type Colombia plants. Upper bands are the PCR fragments specific for At3g55240. Lower bands are the AHA1 PCR fragments used for loading adjustment. Lane 1, leaves; lane 2, stems; lane 3, roots; lane 4, petals; lane 5, seeds. The number of PCR cycles was 40 for At3g55240 and 28 for AHA1.

Although the AtPDH1 we isolated is not the exact counterpart of tobacco VDL it is not difficult to speculate that AtPDH1 plays a role in the chloroplast. To support this idea we found two T-DNA insertion mutants of AtPDH1 that cause albino phenotypes (Figure S4a,b). In these T-DNA insertion lines the expression of AtPDH1 is reduced (data not shown). These results indicate that the native function of a gene identified by the FOX hunting system can be revealed.

F01907 plants not only showed the pale green phenotype, but also showed fast plant development and grew taller than wild-type plants (Figure 3e,f). Since such plant development is reminiscent of that of wild-type plants grown under weak light conditions we designated this phenotype PEL for ‘Pseudo-Etiolation in Light’. The PEL phenotype was caused by the over-expression of an unknown gene At3g55240 (Figure 4b). In contrast to F03024 plants these PEL plants have normal photosynthetic activity (Figure 3h) and relatively normal chloroplast structure except for a lower level of membrane stacking and starch accumulation (Figure S3c). The gene involved could encode a small protein (95 amino acids) with a relatively long 3′-untranslated region (330 bp). We found two homologous sequences from Arabidopsis and also homologous sequences from rice and bean, but not from mammals, indicating that this protein is plant specific (Figures S1c and S2b). When these genes were subjected to the subcellular localization prediction program targetp V1.0 (Emanuelsson et al., 2000), they all showed features characteristic of secretion proteins having an N-terminal transmembrane domain and a neighboring cleavage site. However, no function has been reported for these proteins. Since a knockout line for At3g55240 was not found in the public resource centers, an RNAi construct corresponding to the gene was created and transformed into wild-type Arabidopsis plants. Most of the transgenic plants died at very early stages of development and of those that survived none showed any reduction in the transcription level of the targeted gene (data not shown). Hence, the knockout phenotype of the At3g55240 gene may be lethal. A more detailed characterization of mutants including conditional mutants is required, but the FOX hunting system can give some hints towards the function of unknown or lethal genes.

Expression levels of transgenes and mutant phenotypes

Since FOX lines are produced by ectopic expression of individual Arabidopsis full-length cDNAs it is interesting to know whether there is any correlation between expression levels of the transgenes and the mutant phenotypes.

We examined the expression levels of At3g55240 in several retransformants. Expression levels varied from 1 × 103 to more than 1 × 108 times compared with wild type (Figure 5a). We examined the chloroplast contents and bolting time (leaf number before bolting) of these mutants (Figure 5b,c). We observed a reverse correlation between expression levels of At3g55240 and chlorophyll content and bolting time (Figure 5a–c). These results indicate that the mutant morphology is a consequence of the transgene and that the phenotype varies depending on the transgene expression levels. Interestingly, the transcript was preferentially expressed in rosette leaves in wild-type plants (Figure 4c).

Figure 5.

 Co-segregation of phenotype with expression level of the At3g55240 gene.
(a) Relative expression levels of the At3g55240 gene estimated by real-time PCR.
(b) Relative chlorophyll content of true leaves.
(c) Number of rosette leaves before bolting. The number on the x-axis of each graph indicates individual At3g55240 retransformants in the T2 generation. Plants 19, 21 and 22 showed hygromycin sensitivity indicating they were wild-type segregants.
(d) Growth of individual T2 plants used for these experiments.

Representation of FOX mutants by retransformation with PCR-recovered full-length cDNAs

Considering that the average number of the inserted cDNAs in each FOX line is 2.6 it is very important to define a cDNA that represents the T1 phenotype by retransformation with PCR-recovered full-length cDNAs. To further characterize and examine the relationship between the introduced full-length cDNAs and the mutant phenotypes we selected several FOX mutants from various morphological categories. R00721 and R00745 grew slowly compared with wild type (Figure 6a–c), R01049 had dark green and hyponastic leaves (Figure 6d), R01310 had variegated leaves (Figure 6e), R02347 made flowers at a very early stage of development and had a very compact morphology (Figure 6f) and R02635 was a bushy plant and had weak apical dominance (Figure 6g). After retransformation with the full-length cDNAs isolated by genomic PCR all the original phenotypes were recapitulated. The corresponding genes are summarized in Table 1. It is interesting to know that the genes that give very interesting phenotypes are sometimes caused by over-expression of genes categorized as unknown genes. One of the advantages of our FOX hunting system is that it addresses the function of the genes that are difficult to predicate from homology search or conventional knockout mutation.

Figure 6.

 Phenotype represented after retransformation with individual full-length cDNAs. All plants are in the T1 generation and have been grown for 6 weeks. Representatives of each type are shown.
(a) Non-transformed Arabidopsis wild-type plant.
(b) Retransformant with cDNA isolated from F00721.
(c) Retransformant with cDNA from F00745.
(d) Retransformant with cDNA from F01049.
(e) Retransformant with cDNA from F01310.
(f) Retransformant with cDNA from F02347.
(g) Retransformant with cDNA from F02635.

Table 1.   Recapitulated phenotypes and corresponding genes
LinePhenotypesGeneAnnotations in databasePhysiological function
F00721Slow growthAt3g16400Putative lectin mRNAUnknown
F00745Slow growthAt4g34710Putative arginine decarboxylase SPE2Unknown
F01049Slow growth, dark green and hyponastic leavesAt4g13930Putative hydroxymethyltransferaseUnknown
F01310Slow growth, variegated leavesAt3g25920Putative 50S ribosomal protein L15Unknown
F01907Pale green, early floweringAt3g55240Hypothetical geneUnknown
F02347Dwarf, short internodesAt1g66820Unknown proteinUnknown
F02635Bushy, short internodesAt3g01160Unknown proteinUnknown
F03024Slow growth, pale greenAt1g70070Putative DEAD box helicaseUnknown

Future aspect of our technologies

As an initial experimental procedure we used full-length cDNAs for ectopic expression of transgenes in Arabidopsis in a systematic gain-of-function approach. There have been similar attempts to analyze gene function using conventional non-normalized cDNA libraries introduced randomly into plants (Banno et al., 2001; Jun et al., 2002; LeClere and Bartel, 2001). However, in all cases the cDNAs introduced were not prepared as full-length and not normalized. Over-expression of truncated cDNA is thought to induce more co-suppression of endogenous and exogenous genes than that of full-length cDNA (LeClere and Bartel, 2001). In fact the phenotype described in one of the reports was caused by transcriptional co-suppression of a truncated cDNA (LeClere and Bartel, 2001). Another problem is that without normalization the cDNA species introduced into the plants are biased according to the abundance of cDNAs in the library. It is well known that the most abundant transcripts are from structural, metabolic and housekeeping genes. However, expression of important genes such as transcription factors and genes involved in signal transduction is usually of very low abundance. Therefore only normalized full-length cDNA resources make it possible to analyze gene function comprehensively. The activation tagging technique can enhance gene expression located proximal to the enhancer T-DNA insertion sites and it affects transcription on both sides of the enhancer sequence. Therefore, the mutant phenotype is caused by complex gene expression and it becomes difficult to identify which gene corresponds to the mutant phenotype. The FOX hunting system can eliminate such complexity since the full-length cDNA is expressed in the correct orientation between the CaMV 35S promoter and the NOS terminator. After identification of mutants, genomic PCR is used to isolate the gene that gives rise to the mutant phenotype. These FOX lines will be made available to Arabidopsis researchers through our web site ( and also will be distributed through the RIKEN Bio-resource Center in the future.

Another difference between activation tagging and the FOX hunting system is that the latter expresses only one cDNA species from several splicing variants. That is, the function of any splicing variant can be examined specifically by over-expressing each full-length cDNA. On the other hand, the activation tagging method has its own advantage in that the transcriptional enhancers can enhance the transcription of genes that are not identified or collected as full-length cDNAs, such as micro-RNAs and antisense RNAs.

Making use of the rapidity of our gain-of-functional approach we are also characterizing gene function in rice using rice full-length cDNAs. By using Arabidopsis as the host we are taking advantage of the high transformation efficiency and short generation time of this plant. Once full-length cDNAs have been prepared from any plants species, characterization of gene function can be performed in Arabidopsis and does not depend on the generation time of the native plant. With the determination of the genome sequences of several organisms, including trees and crops, rapid elucidation of individual gene function is desired. We propose that our FOX-hunting system is a rapid and efficient method for elucidating the function of genes identified from these sequencing projects.

Experimental procedures

Complementary DNA mixture comprised normalized full-length cDNAs and internal marker genes

Arabidopsis thaliana Columbia full-length cDNA libraries were constructed using two vectors (Seki et al., 2002): the vector Lambda Zap was used for libraries RAFL2 to RAFL6 and the vector Lambda FLC-1-B (Carninci et al., 2001) was used for libraries RAFL7 to RAFL11. Clones from these libraries were single-pass sequenced to select non-redundant clones and to make a full-length cDNA mixture that comprised about 15 000 cDNA clones at an average final concentration of 14 ng ml−1 for each clone. Since we now know from the RAFL clone full-length sequencing project (Yamada et al., 2003) that nearly one-third of these clones contain one redundant cDNA this clone mixture comprises nearly 10 000 non-redundant fl-cDNA species. The orientation of the full-length cDNA relative to the SfiI cloning sites in the RAFL2 and RAFL3 libraries (corresponding to 1623 clones in total) was opposite to those of the rest of the RAFL clones by accident ( In order to have an internal control for morphological screening four bacterial genes related to auxin synthesis (tms1, iaaM) (Comai and Kosuge, 1982; Klee et al., 1984), auxin sensitivity (rolB) (Furner et al., 1986) and cytokinin synthesis (tmr) (Lichtenstein et al., 1984) were amplified by PCR and cloned into the SfiI site of a pBluescript-derived vector. Each control plasmid was added separately to the full-length cDNA mixture at a concentration of 30 ng ml−1. Deoxyribonucleic acid template and PCR primers used for the amplification of the control genes were as follows: pT281 (tms1), 5′-AGAGGCCAAATCGGCCATGTCAGCTTCACCTCTCCTT-3′, 5′-AGAGGCCCTTATGGCCCTAATTTCTAGTGCGGTAGTTAT-3′; pCP3 (iaaM), 5′-AGAGGCCAAATCGGCCATGTATGACCATTTTAATTCACCC-3′, 5′-AGAGGCCCTTATGGCCCTAATAGCGATAGGAGGCGTTG-3′; pLJ-1 (rolB), 5′-TCCTCTAGAGGCCAAATCGGCCATGGATCCCAAATTGCTATTCCT-3′, 5′-TGATCTAGAGGCCCTTATGGCCTTAGGCTTCTTTCTTCAGGTTTA-3′; pT281 (tmr), 5′-AGAGGCCAAATCGGCCATGGACCTGCATCTAATTTTCG-3′, 5′-AGAGGCCCTTATGGCCCCTAATACATTCCGAACGGATGA-3′.

Agrobacterium library of the normalized full-length cDNAs

The DNA cocktail was digested with SfiI (Takara Bio Inc., Otsu, Japan) and cloned into the SfiI site of an Agrobacterium binary vector pBIG2113SF using T4 ligase (New England BioLabs, Beverly, MA, USA). The pBIG2113SF vector is a derivative from pBIG2113N (Taji et al., 2002) and two SfiI sites are inserted into the XbaI site of pBIG2113N so that the full-length cDNA can be inserted in a sense orientation relative to the 35S promoter (Figure 1a). The ligation was set up with a sixfold molar excess of pBluescript to binary vector. Escherichia coli DH10B (Invitrogen, Tokyo, Japan) was transformed with the ligation product by electroporation and colonies were mixed to isolate a plasmid library. Agrobacterium GV3101 was transformed by electroporation with the plasmid library and the resulting bacterial colonies were mixed to make the Agrobacterium library.

Plant material and harvesting

Transformation and growth of plants are as described elsewhere (Ichikawa et al., 2003). In short wild-type Arabidopsis (Col-0) and the transformed lines were grown at 22°C in a cultivation container system (Arasystem, Gent, Belgium) under long-day conditions (16-h light and 8-h dark). Wild-type plants were transformed by floral dipping (Clough and Bent, 1998) using the Agrobacterium library. Hygromycin-resistant T1 seedlings were selected for 7 days on BAM containing 50 mg l−1 hygromycin and then transferred to soil (Nakazawa and Matsui, 2003). Visible phenotypes were scored and all plants showing phenotypes (FT1P) were transferred to new Arasystem trays and observed further. Either rosette leaves or flowers were harvested from all of the FT1P plants for DNA analysis.

DNA gel-blot analysis and hygromycin resistance testing

Southern blotting was performed as described elsewhere (Meyer et al., 1995). Twenty lines were randomly chosen and T2 plants were grown in soil for 3 weeks as described in ‘Plant material and harvesting’. Leaves from three plants per line were harvested and homogenized in liquid nitrogen. Genomic DNA was isolated using the DNeasy plant mini kit (Qiagen, Tokyo, Japan) according to the instruction manual. A 0.5-kb PCR fragment amplified from pBIG2113SF, containing part of the hygromycin resistance gene, was labeled with a digoxigenin (DIG) DNA labeling mix (F. Hoffmann-La Roche, Basel, Switzerland). The PCR primers for the DNA template were HN (5′-ATGAAAAAGCCTGAACTCACCG-3′) and HC (5′-TCGAGAGCCTGCGCGACG-3′). Hybridization was according to the instruction manual except that it was performed with a probe concentration of 5 ng ml−1 at 44°C using the following solution: 5 × SSC, 50% formamide, 0.1%N-lauroylsarcosine, 0.02% SDS, 2% blocking reagent (F. Hoffmann-La Roche), and that the second washing was performed with a solution of 0.1 × SSC, 0.1% SDS. Chemiluminescence was detected by LumiVisionPRO (Aisin, Kariya, Japan) after the hybridized membrane had been treated with a DIG nucleic acid detection kit (F. Hoffmann-La Roche).

To test for hygromycin resistance, 40 seeds on average from the FT1P lines were germinated on hygromycin-containing BAM medium. Ten days after germination the number of resistant seedlings per germinated seed was scored.

Amplification and cloning of full-length cDNA from FT1P plants

Approximately 200 mg FW of rosette leaves or flowers were harvested. Five ceramic particles (CERAMICS YTZ ball, D: 2.3 mm, Nikkato, Japan) and 300 μl of lysis buffer were added to the plant material to homogenize it using shake master (ver. 1.0, Bio Medical Science, Tokyo, Japan). Genomic DNA was extracted using the Wizard Magnetic 96 DNA Plant System (Promega, Tokyo, Japan). A workstation system (Tecan Genesis Workstation150, Tecan, Tokyo, Japan) was adopted to run the extraction protocol described in the extraction kit. When cloning was into the Gateway vector the following primer pair was used for the cDNA PCR: B1GS7, 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTCTAGAGGCCCTTATGGCCG-3′; B2GS8, 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTCGAGTTAATTAAATTAATCCCCC-3′. When cloning was into the pBIG2113SF vector the following primer pair was used: GS4, 5′-ACATTCTACAACTACATCTAGAGG-3′; GS6, 5′-CGGCCGCCCCGGGGATC-3′. The PCR conditions for short fragments were 94°C for 30 sec for denaturation, 62°C for 30 sec for annealing and 72°C for 120 sec for elongation. The PCR conditions for long fragments were 94°C for 30 sec for denaturation, 58°C for 30 sec for annealing and 68°C for 180 sec for elongation. In both cases the DNA was denatured for 8 min at 95°C prior to the reactions.

Cloning of PCR fragments into the expression vectors and sequencing

The Gateway vector PCR fragments were first cloned into the pDONR-207 vector by BP clonase according to the instruction manual (Invitrogen, Carlsbad, CA, USA) and inserted full-length cDNA fragments were sequenced using the attL1 and attL2 primers: attL1, 5′-TCGCGTTAACGCTAGCATGGATCTC-3′; attL2, 5′-GTAACATCAGAGATTTTGAGACAC-3′. For the pBIG2113SF cloning the PCR fragments from T1 plants were digested with SfiI and cloned into the SfiI site of the vector. The resulting construct was called pF03024 in the case of the F03024 line. Inserted full-length cDNA fragments were sequenced using the GS6 primer.


Seeds of T1 plants from lines F03024 (T1-F03024) and F01907 (T1-F01907), six independent T1-R1907 lines and wild-type plants were sown in soil. F03024 plants were grown for about 6 weeks, and F01907 and R1907 plants for 4 weeks. Rosette leaves from at least three different plants derived from each line were harvested and mRNA was extracted using the Dynabeads mRNA DIRECT kit (Dynal, Oslo, Norway) according to the instructions. For the tissue specificity experiments wild-type Columbia plants were harvested 5 weeks after germination and mRNA from corresponding tissues was isolated in the same way. The mRNA was treated with RQ1Dnase (Promega) for 1 h at 37°C. Complementary DNA was synthesized using the Superscript first-strand synthesis system for RT-PCR (Invitrogen) according to the manufacturer's instructions. Reverse transcriptase-PCR was first performed using the following beta-tubulin specific primers (Takahashi et al., 2001) or Arabidopsis plasma membrane H+-ATPase (AHA1) specific primers (Kinoshita et al., 2001) to adjust the ratio of cDNA between wild type and individual lines: TU1 5′-TTCATATCCAAGGCGGTCAATGTG-3′; TU2, 5′-CCATGCCTTCTCCTGTGTACCAA-3′; AHA1, 5′-TTCTTCTGGGTGAAGATGTCAGG-3′; AHA3, 5′-TGGTTTTAGGAGCAAGACCAGC-3′. Primers specific for the gene in F03024 are: 3024-N, 5′-ATGAACACTCTTCCCGTCGTCTC-3′; 3024-C, 5′-TCAAAGTCTTGCCACTACTAGTCG-3′. Primers specific for the gene in F01907 are: 1907N, 5′-TGATAGAGAAATGTTTGATCTTCCAT-3′; 1907C, 5′-TCTTGCTTGTTGGACCGATGCTAAG-3′. The PCR was always performed under the following conditions: 94°C for 30 sec for denaturation, 60°C for 30 sec for annealing and 72°C for 120 sec for elongation.

Gene expression analysis by quantitative real-time PCR

Ribonucleic acid was isolated from 1-month-old rosette leaves of T2 R01907 plants derived from one T1 R01907 line using a NucleoSpin RNA plant kit (Macherey-Nagel GmbH, Duren, Germany). Complementary DNA was synthesized using Superscript first-strand synthesis system according to the instructions (Invitrogen). Real-time PCR analysis was performed using the MX3000P Multiplex Quantitive PCR System (Promega Corp., Madison, WI, USA). SYBR Green Realtime PCR Master Mix (TOYOBO Co. Ltd, Osaka, Japan) was used for the detection of amplified fragments. Primers for amplification of reference DNA were: ACT2a, 5′-CTGGATCGGTGGTTCCATTC-3′; ACT2b, 5′-CCTGGACCTGCCTCATCATAC-3′. Gene-specific primers for real-time PCR were: RP-1907-2, CATGCGTCAGGGATAAATCGT and LP-1907-2, ACTGTGTGGAAGGAGCTGGA.

Construct for GFP fusion protein and measurement by fluorescence microscopy

The pF03024S clone was used to amplify the DNA fragment corresponding to the 98 amino acids (PDHN98) at the N-terminal end of AtPDH1 using the following primers: attB1-3024N, 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTATGAACACTCTTCCCGTCGTCT-3′; attB2-3024C2, 5′–GGGGACCACTTTGTACAAGAAAGCTGGGTCGTCATCGCTATTCCGAATTTCA-3′. The amplified DNA fragment was cloned into pGWB5 (Saito et al., 1999) through the pDONR-207 vector according to the Gateway instruction manual (Invitrogen). The resulting construct pGWB3024N98 can over-express the 98 amino acids of the N-terminal region fused with the N-terminal end of a synthetic GFP gene (Chiu et al., 1996) under the control of the CaMV 35S transcription promoter. pSA701 was provided by Dr T. Nakagawa (Shimane University, Shimane, Japan) and was used as a negative control construct in which the GFP gene alone was expressed by the 35S promoter. These plasmids were used for particle bombardment of Arabidopsis Col-0 leaves using the Helios Gene-Gun system (Bio-Rad, Tokyo, Japan) following the standard protocol provided by the supplier. Individual leaves were viewed with a fluorescence microscope (BX 60; Olympus, Tokyo, Japan) using the following filters: U-MNIBA for the GFP fluorescence, U-MWIG for chlorophyll autofluorescence.

Electron microscopy

Rosette leaves from 4-week-old plants were fixed in 4% glutaraldehyde, which was buffered with 20 mm sodium cacodylate at pH 7.0 for 20 h at 4°C, and washed with the same buffer for 4 h at 4°C. Then they were post-fixed with 2% osmium tetroxide in 20 mm cacodylate buffer (pH 7.0) for 20 h at 4°C. The fixed samples were dehydrated through an alcohol series and embedded in Spurr's resin (Taab, Berkshire, UK). Ultrathin sections were cut with a diamond knife on an Ultracut UCT ultramicrotome (Leica, Wien, Austria), and transferred to Formvar-coated grids. They were double-stained with 4% uranyl acetate for 15 min and with lead citrate solution for 10 min at room temperature. After washing with distilled water, the samples were observed with a JEM-1200 EX electron microscope (Jeol, Tokyo, Japan) at 80 kV.

Measurements of chlorophyll content

Chlorophyll content was measured in two ways: one method (Figure 3g) was by direct measurement of the pigment as described elsewhere (Porra et al., 1989). In short, leaf material was homogenized in 80% acetone. The acetone solution was measured using a spectrophotometer (Ultrospec 3000, Pharmacia Biotech, Cambridge, UK) at wavelengths of 663, 645 and 720 nm. The second way (Figure 5b) was by determination of the relative content of chlorophyll per unit leaf area using a two-wavelength-type handy chlorophyll meter (SPAD-520; Minolta, Tokyo, Japan). After chlorophyll measurement the same material was used for the quantitative real-time PCR.

Chlorophyll fluorescence measurements

As the quantum yield of photosystem II can be given as Pm′ = (Fm′–Fs)/Fm′ (the Genty parameter), where Fm′ and Fs are maximum and steady-state chlorophyll fluorescence in light-adapted leaves, the chlorophyll fluorescence of leaves of 3-week-old plants was determined at room temperature using a pulse-amplitude-modulated (PAM) fluorometer (MINI-PAM, Walz, Effeltrich, Germany). The results are the means of at least four different leaf measurements.

In silico analysis

The BLAST search was carried out with ncbi blast programs. Subcellular localization was predicted using targetp V1.0 (Emanuelsson et al., 2000) and chlorop 1.1 (Emanuelsson et al., 1999). Multiple alignment was performed by clustalx (Thompson et al., 1997) and genedoc (Nicholas et al., 1997). A neighbour-joining phylogenic tree was constructed using clustalx and treeview (Page, 1996).


We thank Dr M. Kawaguchi (Tokyo University) for providing the following three plasmids: pT281, pCP3 and pT281. We thank Dr T. Nakagawa (Shimane University) for providing the pGWB5 and pSA701 plasmids. We thank A. Enju, J. Ishida, M. Nakajima and M. Narusaka for mixing RIKEN Arabidopsis full-length clones. We thank T. Sakurai and M. Satou for the cDNA sequence clustering. We thank Dr T. Yoshizumi for helping with fluorescence microscopy measurement. This study was carried out during a project on saturation mutagenesis by the Plant Function Genomics Research Group.