A GAL4-based targeted activation tagging system in Arabidopsis thaliana


  • Takamitsu Waki,

    1. Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Search for more papers by this author
  • Shunsuke Miyashima,

    1. Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Current affiliation:
    1. Department of Bio and Environmental Sciences, Institute of Biotechnology, University of Helsinki, Helsinki, Finland
    Search for more papers by this author
  • Miyako Nakanishi,

    1. Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Search for more papers by this author
  • Yoichi Ikeda,

    1. Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Search for more papers by this author
  • Takashi Hashimoto,

    1. Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Search for more papers by this author
  • Keiji Nakajima

    Corresponding author
    1. PRESTO, Japan Science and Technology Agency, Kawaguchi, Saitama, Japan
    • Graduate School of Biological Sciences, Nara Institute of Science and Technology, Ikoma, Nara, Japan
    Search for more papers by this author

  • The authors have no conflict of interest to declare regarding the publication of this paper.

  • Accession number: GenBank accession number for URP1 locus, JX308239.

(e-mail k-nakaji@bs.naist.jp).


Activation tagging is a powerful tool for discovering novel genes that are not easily identified by loss-of-function (lof) screening due to genetic redundancy or lethality. Although the current activation tagging system, which involves a viral enhancer sequence, has been used for a decade, alternative methods that allow organ- or tissue-specific activation are required to identify genes whose strong activation leads to loss of fertility or viability. Here, we established a GAL4/UAS activation-tagging system in Arabidopsis thaliana. Host plants that express a synthetic transcription activator GAL4:VP16 (GV) in an organ- or tissue-specific manner were transformed with a T-DNA harboring tandem copies of UAS, a GAL4-binding sequence. Using a post-embryonic and root-specific GV-expressing line as the host plant, we isolated several dominant mutants with abnormal root tissue patterns, designated as uas-tagged root patterning (urp) mutants, and identified their causal genes. Notably, most URP genes encoded putative transcription factors, indicating that the GAL4/UAS activation tagging system effectively identifies genes with regulatory functions. lof phenotypes of most URP genes were either local patterning defects or visible only if homologous genes were disrupted simultaneously or independently. Systemic overexpression of some URP genes resulted in seedling lethality. These results indicate that GAL4/UAS activation tagging is a powerful method for identifying genes with biological functions that are not readily identified by conventional screening methods.


Forward genetic screening of Arabidopsis thaliana mutants has greatly enhanced our understanding of various biological processes in higher plants. Following the completion of the Arabidopsis genome project, efforts have been made to determine the functions of all annotated genes (Multinational Arabidopsis Steering Committee, 2002). While the collection of T-DNA insertion mutants currently covers 96% of annotated genes, more than two-thirds of Arabidopsis genes remain unassigned to any functional category (Multinational Arabidopsis Steering Committee, 2010). The large gap between the number of insertion mutants and knowledge of gene functions is largely attributable to the fact that a single gene knock-out does not result in a visible phenotype in most cases, due to genetic redundancy (Arabidopsis Genome Initiative, 2000). Although mutant screening of other plant species with a different spectrum of genetic redundancy (e.g. rice) has identified key factors in the biological processes shared by higher plants (Iuchi et al., 2007; Tokunaga et al., 2012), functional assignment of the remaining genes is becoming progressively more difficult and will rely on the invention of new technologies and the improvement of existing ones.

Activation tagging mutagenesis has been used for over a decade to generate dominant mutants for which phenotypic defects can be easily recognized regardless of whether or not the underlying gene exhibits genetic redundancy. In the most widely used system, tandem copies of an enhancer sequence from the cauliflower mosaic virus 35S promoter are randomly inserted into the host genome either by Agrobacterium-mediated transformation or transposon-based mobilization (Weigel et al., 2000; Marsch-Martinez et al., 2002). These sequences enhance the expression of neighboring genes and thereby give rise to a gain-of-function phenotype. While this system has been successfully used to identify some important genes, the number of mutants obtained by this method is far less than anticipated based on the conventional 35S promoter–gene fusion experiments (Weigel et al., 2000). A possible explanation for this discrepancy is that the 35S enhancer affects only transcriptional levels, rather than altering spatiotemporal expression patterns (Weigel et al., 2000). Alternatively, the 35S enhancer sequence might cause strong transcriptional activation in reproductive organs, during gametogenesis, or in embryos, which may lead to infertility or lethality, making isolation of stable mutants impossible for such genes.

Here, we describe a GAL4/UAS activation tagging system that allows the targeted activation of tagged genes in Arabidopsis. In this system, host plants that express a synthetic transcription activator, GAL4:VP16 (GV), in an organ- or tissue-specific manner were transformed with T-DNA harboring an upstream activation sequence (UAS), the binding sequence of GAL4 (Brand and Perrimon, 1993). By using an appropriate GAL4 enhancer trap line as host plant (Haseloff, 1999), one can design a customized targeting strategy. While a similar method has been successfully used in Drosophila (Rorth, 1996), it has hitherto not been used in Arabidopsis. As a model case, we used this method to identify genes that affect the highly organized cellular patterns of the Arabidopsis root meristem (Petricka et al., 2012). Interestingly, most of the genes thus identified encoded transcription factors (TFs) with loss-of-function (lof) phenotypes that were either limited to a certain tissue or became visible only when functionally redundant genes were disrupted simultaneously.


Development of the GAL4/UAS activation tagging system

A basic scheme of GAL4/UAS activation tagging is presented in Figure 1. One can select host plants with the desired GV expression pattern from the collections of GAL4 enhancer trap lines generated by Jim Haseloff (http://www.plantsci.cam.ac.uk/Haseloff/assembly/page138/page138.html) and Scott Poethig (http://arabidopsis.info/CollectionInfo?id=50). Alternatively, transgenic plants that express GV under a promoter with a known expression pattern may be used.

Figure 1.

Scheme of the GAL4/UAS activation tagging system. (a) Structure of the pBIB-UAS tagging vector. Gene fragments are not drawn to scale. (b) Activation tagging strategy used in this study. In the confocal images of roots, green indicates GFP fluorescence, while red delineates root cell walls visualized by the propidium iodide staining. RB, right border; HygR, hygromycin resistance gene; CamR, chloramphenicol resistance gene; LB, left border. Scale bar: 50 μm.

The tagging vector pBIB-UAS harbors five copies of a 17-mer UAS (5×UAS) in the proximity of the T-DNA left border. A polylinker sequence, the plasmid replication origin, and the chloramphenicol resistance gene were incorporated for the purpose of plasmid rescue. A hygromycin resistance gene was used as a selectable marker for plant transformation, so that these lines would be compatible with the kanamycin-resistant GAL4 enhancer trap lines (Figure 1a). We constructed another tagging vector that contained 5×UAS at both ends of the T-DNA to increase tagging efficiency. This plasmid, however, was unstable in Agrobacterium, and hence was not used in this study.

Prior to generating an activation tagging population, we used tagged SHORT ROOT (SHR) to estimate the activity range of 5×UAS with respect to the distance from the tagged gene. SHR is transcribed in the root stele and non-cell-autonomously regulates endodermis differentiation and SCARECROW (SCR) expression in the adjacent cell layer. Ectopic expression of SHR outside the stele led to the formation of supernumerary ground tissue layers (Helariutta et al., 2000; Nakajima et al., 2001). We constructed plasmids in which 5×UAS was placed at various distances from either the 5′ or 3′ end of the SHR coding region (Figure S1a). The constructs were then introduced into the J0571 enhancer trap line, which exhibits ground tissue-specific GV expression in the root (Figure S1b). An analysis of the number of ground tissue layers, as well as the root fasciation phenotype, which we regarded as a strong defects caused by SHR overexpression, indicated that 5×UAS could act within a 0.8-kb region upstream of the first Met codon, but not from further upstream or from the 3′ side, at least in the case of SHR (Figures S1b and c).

Screening of root morphological mutants

We used the GAL4/UAS activation tagging system to isolate mutants with abnormal root morphology (Table 1). As host plants, we used two lines from Haseloff's GAL4 enhancer trap collection, Q2610 and J0571 (Figure 1b). In these lines, expression patterns of GV can be monitored by cell-autonomous GFP fluorescence (Haseloff, 1999). In Q2610, GV is strongly expressed in all root tissues along the entire root length and transiently expressed in progenitor cells of the shoot apical meristem in heart stage embryos, but only weakly expressed or non-existent in other organs (Figures 1b and S2a). This line was expected to activate the expression of tagged genes in all root tissues, while avoiding potential infertility due to the absence of GV expression in the aboveground organs. Another line, J0571, showed highly specific expression of GV in the root ground tissue, along with expression in various organs, including flowers (Figures 1b and S2b).

Table 1. Statistics of the GAL4/UAS activation tagging screening for root morphological mutants
  1. a

    The sole mutant identified in J0571 (urp10-D) was infertile, precluding the analysis of later generations.

Host plantQ2610J0571
Number of T1 lines screened12 0015209
Number of T1 lines with altered root morphology661
Number of lines with their phenotypes reproduced in T2
Patterning defect10a
Cell expansion defect6a
Number of lines with responsible 5×UAS insertion identified
Patterning defect6a
Cell expansion defectNot determineda

The pBIB-UAS tagging vector was introduced into the host plants by Agrobacterium-mediated transformation (Clough and Bent, 1998). After selection on a hygromycin-containing medium, transgenic seedlings were transferred to agar medium that lacked hygromycin to restore root growth. Segregation analysis for hygromycin resistance using T2 progeny from 76 randomly selected lines suggested that 41 lines (54%) harbored a single-locus insertion while 32 lines (42%) harbored insertions at more than two loci (χ2 test, P < 0.05). The remaining three lines (4%) segregated in an abnormal pattern (indicating less than one insertion), presumably due to silencing of the hygromycin resistance gene. A total of 12 001 independent lines was generated in the Q2610 host, and lines showing abnormal root growth phenotypes, such as retarded root growth, thick or thin roots, and wavy growth patterns, were selected (Table 1, Figure 2). We then used confocal laser scanning microscopy (CLSM) to select lines that were primarily affected in root cellular patterns (see Figure 3a,b for the wild-type root cellular pattern). The selected lines were grown on soil and seeds were collected. The root phenotype and segregation patterns were examined in the next generation (T2). Lines showing dominant or semi-dominant segregation ratios for the altered root cellular patterns were selected and collectively named uas-tagged root patterning (urp)1-D to urp7-D. We did not analyze urp6-D further, since the phenotype of this mutant could not be confidently linked to a particular insertion locus (see below). In addition to these patterning mutants, we identified several mutants that exhibited abnormal cell elongation without changes in cell division patterns (Table 1). In this study, we did not analyze these expansion mutants, and focused our analysis on the urp mutants with abnormal cell division patterns.

Figure 2.

Seedlings of homozygous urp mutants 6 days after germination. The host line Q2610 exhibits a wild-type morphology (a), whereas urp mutants exhibit abnormal root growth (b–e). Scale bar: 1 cm.

Figure 3.

Root-specific defects of urp mutants. (a) Schematic representation of the Arabidopsis root tissue pattern (modified from Miyashima et al., 2011). (b)–(i) Confocal images showing cellular patterns in the roots of the Q2610 host line (b), urp mutants (c–h), and a plant recapitulating the urp10-D root phenotype (i). Green signal indicates GFP fluorescence, and red signals delineates propidium iodide-stained cell walls. The GFP signals are not shown in (b)–(h). Arrows point to the region showing patterning defects, while bidirectional arrows indicate relatively large regions of abnormal cellular patterns (see text for details). The inset in (h) is a magnified view of the boxed region. (j) Thirty-one-day-old Q2610 and urp mutant plants. Inset shows a 45-day-old urp7-D plant showing slightly delayed flowering compared with Q2610 and other urp mutants. Scale bar: 50 μm (b–i), 5 cm (j).

In urp1-D and urp2-D, cellular patterns at the region centered at the columella initial cells were affected (arrows in Figure 3c,d). In urp3-D plants, a well-organized cellular pattern typically seen in the Arabidopsis root meristem region was mostly lost (Figure 3e), and post-embryonic root growth was abolished in about a quarter of the T2 progeny. The urp4-D and urp5-D plants exhibited a similar phenotype, i.e. over-proliferation of epidermis and root cap cell layers (arrows in Figure 3f,g). By an unknown mechanism, this defect often occurred asymmetrically around the urp4-D root axis (Figure 3f). In urp7-D, the columella root cap was composed of fewer cell layers, and the region normally occupied by flat columella initials was filled with elongated cells (arrows in Figure 3h). In contrast to the severely affected root morphology, the aboveground parts of each urp mutant were mostly normal and their seed set was comparable to that of the wild type (Figure 3j).

For the J0571 host, a total of 5209 independent lines was generated and individually screened by CLSM for altered GFP expression patterns. Only one line, named urp10-D, was isolated from this screening (Table 1). In urp10-D, cells in the ground tissue layers exhibited ectopic periclinal cell division, resulting in ground tissue composed of four to five cell layers in contrast with the two or three cell layers typically seen in the wild-type root ground tissue (arrows in Figure 3i, shown for a recapitulated line; see below) (Paquette and Benfey, 2005).

Identification of tagged genes

Genomic DNA was prepared from a pool of T2 seedlings for urp1-D to urp7-D, or from the leaves of the primary transformant for urp10-D, which was infertile, and used to determine genomic sequences flanking the inserted 5×UAS. Although the tagging vector had been designed for plasmid rescue, and we could indeed determine the flanking sequences of some urp mutants by plasmid rescue, we found that thermal asymmetric interlaced (TAIL)-polymerase chain reaction (PCR; Liu et al., 1995) was faster and more reliable than plasmid rescue and hence was used routinely to determine flanking sequences of most urp mutants. In some urp mutants, multiple T-DNA insertions were detected (Figure S3). In urp3-D and urp5-D, T-DNA fragments were arranged in a back-to-back configuration with 5×UAS on each T-DNA pointing toward opposite directions from the insertion site (Figure S3). PCR primers were designed for each T-DNA insertion locus and used to test whether the UAS-tagged alleles co-segregated with the root phenotype. By performing this experiment for each insertion locus, co-segregating loci were identified for all mutants (Figure 4). The mutant phenotypes were stably inherited at least up to the T4 generation, except for those of urp4-D, which diminished in the T3 generation.

Figure 4.

Schematic representation of the T-DNA insertion loci responsible for the urp mutant phenotypes. Numbers below the T-DNA insertion sites indicate the distance (bp) from the annotated first Met codon. Thick bar with five triangles, T-DNA with a 59UAS tag.

As a final step to confirm the causal relationship between the tagged genes and the root phenotype, the protein-coding region of each gene was cloned under the 5×UAS-TATA sequence in a binary vector (designated as UAS–URP constructs) and introduced into the same enhancer trap lines as used in the initial screening. For all URP genes, root morphological defects were recapitulated in multiple independent lines (Figure S3). The RNA gel-blot analysis revealed the root-specific accumulation of corresponding transcripts in all urp mutants (Figure S4), except for urp4-D in which the mutant phenotype was diminished, as mentioned above.

The UAS-tagged loci thus confirmed to be responsible for the mutant phenotypes are shown in Figure 4. Consistent with the previous estimation of the 5×UAS activity range using SHR (Figure S1), the 5×UAS tag was inserted within the 0.8-kb region upstream of the annotated first Met codon. The DNA sequence of the urp1-D insertion locus did not match the Arabidopsis genome sequence determined for the Col ecotype (Arabidopsis Genome Initiative, 2000). Subsequent PCR analysis revealed that this sequence occurs in ecotypes C24 (in which Haseloff's enhancer trap lines were made) and Ler, but not in Col or Wassilewskija. A putative open reading frame (ORF) encoding a 230-amino-acid polypeptide was identified in this locus (GenBank accession number JX308239). Because this ORF, but not its frame-shift mutant version, was able to recapitulate the urp1-D phenotype when expressed under the 5×UAS:TATA in Q2610, we concluded that the urp1-D phenotype was caused by the ectopic expression of a protein encoded by this hypothetical ORF (Figure S5).

Most URP genes encode putative transcription factors

All URP genes, except URP1, encoded putative transcription factors (TFs). URP2 encodes an AP2 domain-containing TF PLETHORA3 (PLT3), which acts redundantly with other PLT genes to control root meristem formation and maintenance (Galinha et al., 2007). URP3 encodes a DNA-binding with one finger (DOF) TF, OBF BINDING PROTEIN2 (OBP2), which has been implicated in the regulation of glucosinolate biosynthesis and in influencing auxin levels (Skirycz et al., 2006). Recent reports, however, suggest that OBP2 is involved in root vascular development (Lee et al., 2006; Brady et al., 2011). URP7 encodes a NAM, ATAF1/2 AND CUC2 (NAC) TF, SOMBRERO (SMB), which was previously identified by a marker-based screening of mutants with altered root cap patterning (Willemsen et al., 2008). URP10 encodes a MYB TF, AtMYB64, to which no biological functions have yet been ascribed. Recent microdissection analysis suggests that AtMYB64 is preferentially expressed in female gametophytes (Wuest et al., 2010). URP4 and URP5 encode members of the RWP-RK DOMAIN (RKD) TF family, RKD2 and RKD1, respectively (see below) (Schauser et al., 2005).

To examine normal expression patterns of URP genes, we isolated the 5′ upstream region of each URP gene and fused it to a chimeric reporter, nls:YFP:GUS (nYG) (Waki et al., 2011) (see Table S1 for promoter length). The resulting constructs were transformed into wild-type Arabidopsis. Multiple independent lines were analyzed by histochemical GUS staining, as well as CLSM observation of YFP fluorescence. Except for URP5/RKD1, which did not yield a consistent expression pattern, all URP genes were expressed in a subset of tissues (Figures 5 and S6). Reporter expression from the URP4/RKD2 promoter was low, with weak GUS staining barely detectable in the hydathode and stipule (Figure S6d). Expression patterns of URP2/PLT3, URP3/OBP2, and URP7/SMB were consistent with previous reports. URP2/PLT3 showed a graded expression pattern that peaked around the quiescent center (QC) and decreased toward the upper (proximal) region in the central vascular cylinder (Figures 5b, g and S6h) (Galinha et al., 2007). URP3/OBP2 was preferentially expressed in the phloem and associated cell files in roots, as well as in vascular cells in cotyledons (Figures 5c,h and S6c,i) (Lee et al., 2006). URP7/SMB was expressed exclusively in differentiated root cap cells, both in the columella and lateral root cap lineages (Figures 5d, i and S6k) (Willemsen et al., 2008).

Figure 5.

Analysis of URP expression patterns using promoter-nls:YFP:GUS reporter lines. (a)–(e) Confocal laser scanning microscope observation of YFP fluorescence. (f)–(j) Cross-sections from the meristematic zone of the roots of reporter lines stained for GUS activity. Triangles point to the protophloem cell files. Inset in (i) is a section at the distal root tip composed solely of root cap cells. Scale bars: 100 μm (a)–(e), 10 μm (f)–(j).

Further characterization of selected URP mutants and related genes

We noticed that the urp7-D roots were hairless (Figure 6a,b). Observation of root cross-sections revealed that the mature region of urp7-D roots lacks the epidermis layer (Figure 6c, compare with 6e). To examine whether ectopic expression of SMB affected the differentiation status of root tissues, we performed a marker analysis. Because urp7-D roots expressed GFP derived from the enhancer trap locus, we generated a GFP-free transgenic line that expresses SMB under the control of the dexamethasone (DEX)-inducible GAL4:VP16:GR (GVG) transcription factor (Aoyama and Chua, 1997) that in turn was regulated by the Q2610 enhancer region (designated as Q2610 promoter, Figure S7). The resulting plants (UAS-URP7 Q2610pro-GVG, designated Q2610-iSMB) recapitulated the urp7-D phenotype when grown in the presence of DEX. We then crossed Q2610-iSMB with pGL2-GFPer and pCo2-YFPH2B plants that mark non-hair epidermis and cortex cells, respectively (Masucci et al., 1996; Heidstra et al., 2004). The results indicated that ectopic SMB expression eliminated the epidermal cell layer without affecting the differentiation of the cortical layer (Figure 6g,h).

Figure 6.

Role of URP7/SMB in root cap differentiation. (a), (b) A urp7-D primary root exhibiting a hairless phenotype. (b) A magnified image of the region enclosed in a red box in (a). The inset in (b) shows a Q2610 root, which has normal root hairs. (c)–(f) Root cross-sections from the differentiation zone (c, e) and meristematic zone (d, f) of urp7-D (c, d) and Q2610 (e, f) plants. Red triangles indicate ectopic anticlinal divisions in urp7-D epidermal cells. The approximate regions used are indicated in panel (a). The number of epidermal cell files per root circumference [±standard deviation (SD), n = 6] are shown below panels (d) and (f). (g), (h) Root of the dexamethasone (DEX)-inducible root-specific SMB-over-expressing plants (Q2610-iSMB) harboring pGL2-GFPer (g) and pCo2-YFPH2B (h) markers. Plants were grown in the presence of DEX. (i), (j) Root cross-sections of wild-type Col (i) and smb-3 (j) plants, showing fewer lateral root cap (LRC) cell files in smb-3. The number of LRC cell files per root circumference (±SD, n = 8 or 12) is shown below panels (i) and (j). (k), (l) The DEX-inducible SMB-overexpressing seedlings (a systemic overexpression line, 35S-iSMB,) grown in the absence (k) or presence (l) of DEX. En, endodermis; Co, cortex; Ep, epidermis; LRC, lateral root cap. Scale bar: 200 μm (a); 50 μm (b–h); 20 μm (i, j); 5 mm (k, l).

In the meristematic region of urp7-D roots, cells at the epidermis position exhibited planes of anticlinal cell division, resulting in the formation of more cell files around the root circumference (red triangles in Figure 6d, compare with 6f). In Arabidopsis roots, adjoining epidermis and lateral root cap (LRC) cells are derived from a single initial cell. The daughter cells destined to become LRC cells undergo more rounds of anticlinal cell division than those destined to become epidermis (Wenzel and Rost, 2001), making LRC cell files thinner than epidermal cell files. The altered cell division patterns observed in the upr7-D epidermis suggest that URP7/SMB has the capacity to promote anticlinal cell divisions that normally take place in the early phase of LRC differentiation. Consistent with this view, the LRC of lof smb-3 mutants was composed of fewer and thicker cell files than was that of the wild type (compare Figure 6i and j). These observations suggest that SMB promotes early LRC patterning, in addition to its postulated roles in columella patterning (Willemsen et al., 2008).

We also constructed 35S-iSMB (UAS-URP7 35S-GVG) plants that systemically express SMB in a DEX-dependent manner. When grown in medium containing DEX, growth of the whole seedlings was severely affected and eventually terminated (Figure 6l; compare with the non-induced control seedlings shown in Figure 6k). A similar result was reported previously (Bennett et al., 2010). This observation demonstrated that the root-specific activation tagging strategy was efficient at isolating a viable mutant that ectopically expresses SMB.

We were intrigued by the conspicuous phenotype of urp4-D and urp-5D. Both mutants showed a dramatic proliferation of epidermis and root cap cells on hormone-free medium, suggesting that URP4/RKD2 and URP5/RKD1 potentially had important roles in the control of cell division (Figure 3f,g). We expressed GFP-fused RKD1 and RKD2 under the Q2610 promoter (Figure S7). For both RKD1 and RKD2, GFP fluorescence was detected exclusively in the nucleus, as expected from the postulated role of RKD proteins as TFs (Figure 7a,b) (Schauser et al., 2005; Koszegi et al., 2011). The cell proliferation phenotype was recapitulated in RKD1/2-GFP transgenic roots (Figure 7a,b). Furthermore, we introduced UAS-RKD2 constructs into 35S-GVG plants (35S-iRKD2). In the presence of DEX, the seedlings were transformed into a callus-like mass of cells (Figure 7c), suggesting that RKD2 has the ability to activate cell division not only in the root cap cells but also in a broader range of cell types. This was further corroborated by crossing 35S-iRKD2 with Cyclin B1;1-GUS plants, that marks actively dividing cells (Colon-Carmona et al., 1999); GUS-positive cells were detected in most root tissues and extended toward the basal region (Figure 7d). This result further indicates that urp4-D could be isolated by the root-specific activation tagging strategy adopted in this study.

Figure 7.

URP5/RKD1 and URP4/RKD2 have a capacity to activate cell proliferation. (a), (b) Confocal laser scanning microscope images of roots expressing GFP-tagged RKD1 (a) and RKD2 (b). (c) A seedling systemically expressing RKD2 (35S-iRKD2). (d) The expression of the CyclinB1;1-GUS marker extends toward the basal region of the 35S-iRKD2 root after dexamethasone (DEX) induction. The bracket indicates the distribution of GUS-positive cells focused at the meristematic region typically observed for the wild-type root. Scale bar: 20 μm (a, b); 1 mm (c); 200 μm (d).

To analyze the functions of RKD1 and RKD2 in normal development, we obtained T-DNA insertion lines for RKD1 and RKD2 (Figure S8). However, we could not detect morphological defects in either of the homozygous insertion lines or in their double mutants as described previously (Koszegi et al., 2011). Since none of the five RKD genes in the Arabidopsis genome have been assigned with biological functions at the time of our analysis, we extended our lof analysis to the remaining three RKD genes, RKD3, RKD4, and RKD5 (Figure S8). This approach led us to identify RKD4 as a key regulator of embryonic pattern formation. We have published a comprehensive analysis of RKD4 functions separately (Waki et al., 2011).


Activation tagging is a powerful method for identifying genes with lof mutations that do not result in an obvious phenotype due to genetic redundancy. A system based on the 35S enhancer tag has been used successfully (Weigel et al., 2000). As mentioned previously, integration of the 35S enhancer sequence into the Arabidopsis genome probably affects the level of gene expression rather than causing ectopic expression (Weigel et al., 2000). This is advantageous if the gene products act in a dose-dependent manner. However, when screening for patterning genes, enhancement of gene expression levels may not lead to a strong phenotype, and the biological functions of such genes may be effectively visualized when gene expression patterns are altered. This limitation may be partially complemented by screening the FOX-hunting population, which consists of transgenic lines harboring about 10 000 full-length cDNA clones driven by the 35S promoter (Ichikawa et al., 2006). This population, however, is unlikely to include genes whose systemic activation leads to lethality and/or infertility.

In this study, we used two GAL4 enhancer trap lines, Q2610 and J0571, as host plants (Haseloff, 1999). In Q2610, GV expression is largely confined to the post-embryonic roots. By screening 12 001 transgenic lines generated in this background, we identified six mutants with a confirmed 5×UAS insertion (Table 1). This efficiency is comparable to the conventional 35S-based system (Weigel et al., 2000), but less than that reported for a GAL4-based system in Drosophila (Rorth, 1996). This is probably because the Drosophila vector included a minimal promoter (transcription start site), and because P-element-based mobilization was adopted in the Drosophila screening (Rorth, 1996).

While all of the urp mutants identified in the root-specific Q2610 host were viable and fertile, the induced systemic expression of URP genes by the 35S promoter resulted in lethality or infertility (Figure 6l), (Bennett et al., 2010; Koszegi et al., 2011; Krizek and Eaddy, 2012). Nevertheless, in all cases tested (OBP2, SMB, and RKD2), overexpression of each URP gene resulted in the same root phenotype, regardless of whether the gene was expressed in the Q2610 enhancer trap line or the 35S-GVG plant. These observations demonstrate that the use of Q2610 plants as host successfully restricted the overexpression phenotype to the root, allowing the isolation of dominant mutants that could not have been obtained by the conventional system. In contrast, use of J0571 as a host plant allowed us to isolate only one mutant (urp10-D), which was infertile even in the primary transformant. We nevertheless identified the tagged gene as AtMYB64 by amplifying the flanking genomic sequence from the primary transformant and recapitulating the mutant phenotype by transforming UAS-AtMYB64 into J0571. Since AtMYB64 is preferentially expressed in the female gametophytes (Wuest et al., 2010), the GV-dependent ectopic expression of AtMYB64 in J0571 floral organs (Figure S2b) was a likely cause of the infertility. We recently generated a DEX-inducible version of the J0571 line (Miyashima et al., 2011). Use of this line as a host should avoid potential infertility and/or embryo-lethality problems caused by the tagged genes.

One consideration to be taken into account when interpreting an activation tagging phenotype, and indeed any overexpression phenotype, is whether or not the observed phenotype reflects the gene's function in its normal expression site. Therefore we propose that, upon identification of a potentially interesting gene, information about its normal expression pattern be retrieved from publicly available expression data, followed by careful analysis of the corresponding lof mutant. Analyzing the lof mutants may seem paradoxical; however, careful examination of the tissues in which the gene is normally expressed may reveal subtle morphological changes that may be overlooked in a simple lof screening. Using this strategy, we identified URP7/SMB as a regulator of root cap differentiation. The epidermis of urp7-D exhibited a root cap-like cell division pattern, whereas the opposite phenotype, i.e. delayed LRC division, was found in a lof allele of URP7/SMB (Figure 6). Ben Scheres and co-workers identified SMB as a regulator of root cap cell division by marker-based screening (Willemsen et al., 2008), and more recently reported its involvement in root cap maturation (Bennett et al., 2010).

In cases in which no defects can be detected in lof alleles, the analysis has to be extended to homologous genes. This is exemplified by our ongoing characterization of urp3-D. The T-DNA insertion mutant of URP3/OBP2 did not exhibit morphological defects at the site of its expression, the root stele. However, overexpression of OBP2 fused with a transcription repression domain (Hiratsu et al., 2003) dramatically reduced the number of stele cell files, a phenotype opposite to that of urp3-D (our unpublished results). This observation indicates that OBP2 and its functionally redundant genes are involved in the proliferation of root stele cells. A detailed characterization of OBP2 and its homologs is under way. Similarly, in the case of urp4-D and urp5-D, systematic lof analysis of the RKD gene family led us to identify RKD4 as a key regulator of early embryogenesis (Waki et al., 2011).

In conclusion, we have established a GAL4-based targeted activation tagging system in Arabidopsis. Application of this system in both developmental and physiological studies, coupled with its further improvement, e.g. transposon-based mobilization of the UAS tag, use of DEX-inducible GVG driver lines as a host, and marker-based screening to identify regulators of specific genes or events of interest, will boost the rate of functional characterization of unexplored genes in the Arabidopsis genome.

Experimental procedures

Plant materials

Enhancer trap lines generated by Dr Jim Haseloff and co-workers (ecotype C24) were obtained from the Arabidopsis Biological Resource Center (ABRC). smb-3 (SALK_143526) and pCo2-YFPH2B were gifts from Dr Ben Scheres (Heidstra et al., 2004; Willemsen et al., 2008). rkd1-2 (SALK_089683), rkd2-2 (GABI_237C07), rkd4-1 (INRA_331F10), and rkd4-2 (INRA_301B07) were described previously (Koszegi et al., 2011; Waki et al., 2011). Other T-DNA insertion lines used are as follows: rkd1-3 (SALK_061813), rkd2-4 (SAIL_94B05), rkd3-1 (SALK_020897), rkd3-2 (SALK_151524), rkd3-3 (SALK_032952), rkd5-1 (SALK_024414), and rkd5-2 (SALK_055555).

Vector construction

Primers used for vector construction are listed in Table S1.

UAS-SHR constructs

A 4.0-kb AvrII–EcoRI fragment that included the 1.6-kb entire coding region of SHR and the 2.4-kb 5′ upstream region was subcloned from a bacterial artificial chromosome (BAC) clone F18N21 into the SpeI–EcoRI site of the pBluescript II SK(–) vector to give pBS-4.0k-SHR. A DNA fragment containing 5×UAS was amplified from pBINYFPAEQ (Kiegle et al., 2000) using the primers SacII-Bam–UAS5′ and NotXho–UAS3′, and inserted into the SacII–NotI site of pBS-4.0k-SHR to give pBS-UAS-4.0k-SHR. Serial deletion of the SHR promoter was performed by digesting pBS-UAS-4.0k-SHR with XhoI at the 3′ side of 5×UAS, together with the following restriction enzymes present in the SHR promoter: NdeI (for a 865-bp promoter with respect to the first ATG codon), AccI (for a 312-bp promoter), and XbaI (for a 167-bp promoter). Digested plasmids were self-ligated after blunt-end treatment using T4 DNA polymerase. For insertion of 5×UAS into the 3′ side of SHR, a 0.6-kb fragment containing a C-terminal coding region and the 362-bp 3′ region of SHR was amplified by PCR from Col genomic DNA, with the primers SHR-3751 and Apa-SHRter3′, and digested with EcoRI and ApaI. The resulting 0.3-kb fragment was inserted into the EcoRI–ApaI site of pBS-4.0k-SHR to reconstitute a fragment that included a 2.4-kb 5′ upstream sequence, 1.6-kb coding region, and 0.4-kb 3′ region of SHR to give pBS-4.4k-SHR. A DNA fragment containing 5×UAS was amplified from pBINYFPAEQ using the primers Kpn-UAS5′ and EcoRI–ApaI-UAS3′. This fragment was inserted into the EcoRI–KpnI site of pBS-4.0k-SHR to place the 5×UAS 31-bp 3′ from the SHR stop codon. The same 5×UAS fragment was inserted into the ApaI–KpnI site of pBS-4.4k-SHR to place the 5×UAS 362-bp 3′ from the SHR stop codon. The UAS-SHR fragments thus constructed were inserted into the pBIB-KS (Waki et al., 2011) binary vector either as an XhoI–KpnI or a BamHI–KpnI fragment.

pBIB-UAS tagging vector

A DNA fragment containing 5×UAS-TATA was amplified from pBINYFPAEQ (Kiegle et al., 2000) with the primers SalBgl-UAS5′ and BamBstAfl-UAS3′, digested with SalI and BamHI, and inserted into the SalI–BamHI site of the pHSG399 vector (Takara Bio, http://www.takara-bio.com/index.htm) to give pHSG399-UAS. A DNA fragment encompassing a poly-linker sequence, a replication origin, a chloramphenicol resistance gene, and a 5×UAS-TATA sequence was excised from pHSG399-UAS and inserted into pBIB-KS (Waki et al., 2011) as a BstBI–AflII fragment to give pBIB-UAS-TATA. The TATA-box sequence flanking the 5×UAS was removed from pBIB-UAS-TATA by digestion with XbaI and AflII, and the vector was blunted by Klenow treatment and self-ligated to give pBIB-UAS.

UAS-URP constructs

For UAS-URP1 the putative coding region in the insertion flanking sequence of urp1-D was amplified from C24 genomic DNA with the primers Xho–URP1-ORF-Nt and EcoRI-URP1-ORF-Ct, and inserted into the XhoI–EcoRI site of pBIB-UAS-NosT (Waki et al., 2011). A frame-shifted version was constructed in the same way, except that the primer Xho–URP1-ORF-Nt-FS was used in place of Xho–URP1-ORF-Nt.

For UAS-URP10 a DNA fragment containing the AtMYB64 coding sequence and introns was amplified from Col genomic DNA with the primers Apa–MYB64-Nt-F and Xba–MYB64-Ct-R, and inserted into the XhoI–EcoRI site of pBIB-UAS-NosT (Waki et al., 2011).

For UAS-URP2/3/4/5/7 a UAS-tagged gene fragment was amplified from the genomic DNA prepared from the corresponding urp mutants. The primers used were UAS-TAIL3 (anneals to the 5′ side of inserted 5×UAS) and 3′ primers designed for each URP gene (Table S1). The PCR products were inserted into the SalI–BamHI site of pBIB-KS (Waki et al., 2011).

Q2610pro-RKD1/2:GFP constructs

The T-DNA insertion site in the Q2610 enhancer trap line was identified by TAIL-PCR (Liu et al., 1995) using GAL4-VP16-1/2/3 primers. A 3.0-kb genomic fragment upstream of the insertion site (Figure 7a) was amplified from genomic DNA prepared from Q2610 with the primers Kpn–At5g48110-NtR and Apa–GAL4-ProEnd, and inserted at the KpnI–ApaI site of pBIB-GFP-NosT (Waki et al., 2011) to give pBIB-Q2610pro-GFP-NosT. RKD1- and RKD2-coding sequences were amplified by reverse-transcription (RT)-PCR from RNA prepared from UAS-RKD1/35S-GVG and UAS-RKD2/35S-GVG transgenic plants, respectively, grown in the presence of the DEX inducer. Primers used were Apa-RKD1-5′ and Bam-RKD1-Gly-3′ for RKD1, and Apa-RKD2-5′ and Bam-RKD2-Gly-3′ for RKD2. The PCR product was inserted between the Q2610 promoter and NosT of pBIB-Q2610pro-GFP-NosT as an ApaI–BamHI fragment.

Q2610pro-GVG and 35S-GVG constructs

The GVG-coding sequence was obtained by fusing a GAL4:VP16 (GV)-coding sequence from pBIN-GV (Waki et al., 2011) and a rat glucocorticoid receptor (GR)-coding sequence from pBI-ΔGR (Lloyd et al., 1994), and used to replace the GUS-coding region of pBI121 (Clontech-Takara Bio, http://www.takara-bio.com/) to give pBIN-35S-GVG. A DNA fragment harboring the UAS-GFPer-NosT gene cassette was amplified from pBIN-UAS-GFPer-NosT (Waki et al., 2011) with the primers Nhe-UAS-5′ and NheHind–NosT-3′, digested with NheI, and partially filled-in with dCTP and dTTP. This fragment was inserted at a HindIII site of pBIN-35S-GVG that had been partially filled-in with dATP and dGTP, to give pBIN-UAS-GFPer-35S-GVG. For construction of Q2610pro-GVG, the Q2610 promoter fragment described above was digested with HindIII and BamHI, and used to replace the 35S promoter of pBIN-35S-GVG to give pBIN-Q2610pro-GVG.

Promoter-nls:YFP:GUS constructs

The 5 upstream region of each URP gene was amplified from genomic DNA prepared from the wild type C24 (for URP1) or Col (for other URPs) with the primers listed in Table S1. The PCR products were digested with the restriction enzymes incorporated at the 5′ ends of the primers and inserted into pBI-Kan-nlsYG (Waki et al., 2011). The length of the promoter regions can be found in Table S1.

GL2pro-GFPer construct

The endoplasmic reticulum-targeted GFP-coding sequence was amplified from pBIN-35S-GFPer (a gift from Dr Jim Haseloff) (Haseloff, 1999) with the primers Hind–GFPer-Nt and Xho–GFPer-3′R and cloned into the HindIII–XhoI site of pBluescript II SK(–) (Agilent Technologies, http://www.home.agilent.com/) to give pBS-GFPer. A polyadenylation sequence derived from Agrobacterium tumefaciens Gene 7 (Velten and Schell, 1985) was amplified from the pBIB vector (Becker, 1990) with the primers Xho–Gene7-ter-F and Kpn–Gene7-ter-R and inserted downstream of the GFPer-coding sequence of pBS-GFPer to give pBS-GFPer-Gene7ter. A 2.0-kb GL2 promoter was amplified from the genomic DNA prepared from A. thaliana ecotype Columbia with the primers Bam-GL2-(–)2047 and Hind–GL2pro-End-R and inserted upstream of the GFPer-coding sequence of pBS-GFPer-Gene7ter to give pBS-GL2pro-GFPer-Gene7ter. The GL2pro-GFPer-Gene7ter fragment was excised with BamHI and KpnI, and inserted into the pBI-Kan binary vector (Waki et al., 2011) to give pBI-Kan-GL2pro-GFPer.

Identification of tagged loci

Genomic DNA was extracted from Arabidopsis seedlings using a buffer composed of 0.1 m 2-amino-2-(hydroxymethyl)-1,3-propane diol (TRIS)-HCl pH 8.0, 1.4 m NaCl, 20 mm EDTA and 2% (w/v) CTAB, extracted once with chloroform/isoamylalcohol (24:1), and precipitated with isopropanol. Genomic fragments flanking the UAS insertion sites were amplified by TAIL-PCR (Liu et al., 1995) using the primers listed in Table S1. The PCR products were purified on QIAquick Columns (Qiagen, http://www.qiagen.com/) and used as templates for sequencing by BigDye Terminator version 3 (Life Technologies, http://www.lifetechnologies.com/) with the primer UAS-TAIL3.

Expression analysis

For RNA gel blot analysis, total RNA was extracted from the shoots and roots of 12-day old seedlings using an RNeasy Plant Mini Kit (Qiagen). For probe preparation, cDNA fragments of URP genes were amplified by PCR using the primers listed in Table S1, cloned into the pGEM-T vector (Promega, Madison, WI, http://www.promega.com), and radiolabelled with 32P. The GUS staining, tissue sectioning, and microscopic observation were performed as described previously (Nakajima et al., 2004; Miyashima et al., 2009).


Tissue sections were observed using a Nikon Eclipse E1000 Microscope (Nikon, http://www.nikon.com). CLSM was performed using a Nikon C1 Confocal Laser Scanning Microscope. Roots were stained with 7 μg/ml propidium iodide before observation.


We are grateful to Dr Jim Haseloff and Dr Edward Kiegle for providing GAL4 enhancer trap lines and plasmid vectors. We thank Naoko Nishida, Hiroe Ohnishi and Yuko Ono for generating the transgenic plants. This work was supported by grants from the Japan Society for Promotion of Science (JSPS) (15770146, 18510171, 20061022 and 21027025), the Novartis Foundation (Japan) for the Promotion of Science (16-198), and The Sumitomo Foundation (060507). S.M. was supported by JSPS Research Fellowships for Young Scientists.