Gene silencing induced by hairpin or inverted repeated sense transgenes varies among promoters and cell types

Authors

  • Gordana Marjanac,

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author
  • Mansour Karimi,

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author
  • Mirande Naudts,

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author
  • Tom Beeckman,

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author
  • Anna Depicker,

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author
  • Sylvie De Buck

    1. Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium
    2. Department of Plant Biotechnology and Genetics, Ghent University, 9052 Gent, Belgium
    Search for more papers by this author

Author for correspondence:
Ann Depicker
Tel: +32 93313940
Email: ann.depicker@psb.vib-ugent.be

Summary

  • In transgenic calli and different tissues of Arabidopsis thaliana plants, the in trans silencing capacity of a 35S-β-glucuronidase (GUS) hairpin RNA construct was investigated on a target GUS gene, under the control of the 35S, a WRKY or several cell cycle-specific promoters.
  • GUS histochemical staining patterns were analyzed in all tissues of the parental lines and supertransformants harboring the hairpin construct. Quantitative GUS activity measurements determined GUS suppression by a 35S-GUS hairpin or inverted repeated GUS transgenes in leaves and calli.
  • In some supertransformants, GUS-based staining disappeared in all tissues, including calli. In most supertransformants, however, a significant reduction was found in mature roots and leaves, but residual GUS activity was observed in the root tips, young leaves and calli. In leaves of most hairpin RNA supertransformants, the GUS activity was reduced by c. 1000-fold or more, but, in derived calli, generally by less than 200-fold. The silencing efficiency of inverted repeated sense transgenes was similar to that of a hairpin RNA construct in leaves, but weaker in calli.
  • These results imply that the tissue type, nature of the silencing inducer locus and the differential expression of the targeted gene codetermine the silencing efficiency.

Introduction

In plants, the expression of transgenes and endogenes can be reduced as a result of homology-dependent gene silencing on transcriptional (TGS) and/or post-transcriptional (PTGS) levels (Baulcombe, 2004; Matzke & Birchler, 2005; Meins et al., 2005; Wang & Metzlaff, 2005; Eamens et al., 2008; Kanazawa, 2008). Both types of silencing are frequently associated with the presence of direct or inverted repeated sense T-DNA transgenes, but single-copy or multiple transgenes that are not arranged in inverted repeats can also induce effective silencing (Muskens et al., 2000; De Buck et al., 2001; De Wilde et al., 2001; Lechtenberg et al., 2003). Since the discovery that double-stranded RNA (dsRNA) is responsible for silencing induction (Fire et al., 1998; Hamilton & Baulcombe, 1999), various expression vectors for an intron-spliced, self-complementary hairpin RNA (hpRNA) have been designed to obtain efficient PTGS in plants (Helliwell et al., 2002; Dafny-Yelin et al., 2007; Karimi et al., 2007).

RNA-induced gene silencing, also called RNA interference (RNAi), is based on the delivery of dsRNA to effectively silence the targeted gene by a sequence-specific RNA degradation mechanism (Brodersen & Voinnet, 2006; Vaucheret, 2006; Small, 2007). The dsRNA is processed into small interfering RNAs (siRNAs) of 21–28 nucleotides in length by an RNase III-like enzyme (Dicer) (Bernstein et al., 2001). Subsequently, these siRNAs are incorporated into an RNA-induced silencing complex to guide the degradation of homologous mRNAs in the cytoplasm, DNA methylation of homologous sequences and suppression of translation (Van Houdt et al., 1997; Dunoyer & Voinnet, 2005; Vaucheret, 2006; Brodersen et al., 2008).

The silencing efficiency by which a silencer locus can downregulate the expression of a target gene seems to be strongly linked to the levels of dsRNA and siRNA production. Indeed, a hairpin construct controlled by the strong 35S promoter induces a stronger silencing phenotype than the same construct controlled by a weak nopaline synthase promoter (Chuang & Meyerowitz, 2000). In addition, inverted repeated sense transgenes show a positive correlation between the silencing efficiency and zygosity of the silencing-inducing locus, as well as of the primary target locus (García-Pérez et al., 2004; Bleys et al., 2006a,b).

Although the expression of both endogenes and transgenes can be suppressed by hpRNA transgenes, it is unclear whether hpRNA-mediated gene suppression can silence genes in all cell types equally. In particular, the downregulation of genes in tissues containing highly proliferating cells has been questioned (Mitsuhara et al., 2002; Teerawanichpan et al., 2004). Moreover, different conclusions have been reported on the silencing capacity of inverted repeated sense transgenes and even single-copy genes in tissues containing highly proliferating cells. For example, both PTGS and DNA methylation were released in calli induced from leaves displaying silencing of both inverted repeated and single-copy sense transgenes (Guo et al., 1999; Mitsuhara et al., 2002; Corrêa et al., 2004). By contrast, transgene silencing of a single-copy gene was released only after RNAi-mediated inactivation of DNA methyltransferases (Teerawanichpan et al., 2004). Furthermore, an epigenetic switch from PTGS to TGS in inverted repeated sense transgenes occurred 24 months after in vitro cultivation (Fojtova et al., 2003).

Until now, the silencing capacity of a silencing locus has not been compared in different tissues and meristematic cells. To investigate whether hpRNA-induced silencing can be generated equally efficiently in all plant cell types, two experimental procedures were followed. In a first approach, we determined the efficiency of hpRNA-mediated β-glucuronidase (GUS) transgene silencing in plants expressing the GUS reporter gene under the control of the cauliflower mosaic virus (P35S), B-type cyclin CYCB1;1, D-type cyclin CYCD4;1, Kip-related protein KRP4 and WRKY23 promoters (see Materials and Methods section). In a second approach, we examined whether the in trans silencing in leaves, induced by inverted repeated GUS sense genes or by a hpRNA GUS construct, could be maintained in transgenic Arabidopsis calli induced from these leaves.

Materials and Methods

Plasmids and Agrobacterium strain

The 792-bp fragment of the 3′-end of the GUS coding sequence was recombined into the destination vector pH7GWIWG2 (Karimi et al., 2002) to produce the expression clone pHhpUS. This 792-bp fragment was amplified by PCR with the pK2L610 plasmid (De Buck et al., 1998) as template and two sets of primers to form the attB1 and attB2 recombination sites. In the first step, template-specific primers containing 12 bases of the attB sites were used in 30 cycles of PCR to amplify the target sequence. The PCR conditions were as follows: initial denaturation at 94°C for 5 min, followed by 30 cycles of denaturation at 94°C for 1 min, annealing at 55°C for 1 min and elongation at 68°C for 90 s, and a final elongation step at 68°C for 15 min. This product was subsequently used as a template in the second PCR with universal attB adapter primers to amplify the full attB1 and attB2 recombination sites (Invitrogen, Carlsbad, CA, USA), following the protocol described in the instruction manual. The primers used were forward template-specific primer 5′-AAAAAGCAGGCTTGCTGGACTGGGCAGATGAA-3′, reverse template-specific primer 5′-AGAAAGCTGGGTTTGCCTCCCTGCTGCGGTTT-3′, attB1 adapter primer 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCT-3′ and attB2 adapter primer 5′-GGGGACCACTTTGTACAAGAAAGCTGGGT-3′.

To obtain the 3′ CS-GUS entry clone, this 792-bp PCR product, flanked by attB sites, was recombined into the pDONR201-KmR vector, containing attP1 and attP2 recombination sites, using the BP Clonase enzyme (Invitrogen). Finally, the expression vector pHhpUS (Fig. 1) was generated in the LR reaction, during which the 3′ CS-GUS entry clone was incubated with the pH7GWIWG2(I) destination vector in the presence of the LR Clonase enzyme (Invitrogen). The orientation of the intron after double LR reaction was determined by restriction digests. The transformation vector pHhpUS was transformed into the Agrobacterium strain C58C1RifR, containing the pMP90 vir plasmid (Koncz & Schell, 1986).

Figure 1.

 Schematic representation of the pHhpUS vector. (a) T-DNA expression vector pHhpUS generated in an LR reaction with the 3′ CS-GUS entry clone and the GATEWAY-compatible binary T-DNA destination vector, pH7GWIWG2(I) (Karimi et al., 2002) (see Materials and Methods section). The attB3′ CS-GUS cassette is inverted repeated in the hairpin. The intron will be spliced out when the hairpin construct is expressed in plants. (b) Schematic outline of the T-DNA locus structure in the transformants FK24, KH15, KH15d6 and KH9/2 and indication of the structural elements of the K and H T-DNAs. The transgene locus in FK24 consists of the K T-DNA, whereas the transgene locus in line KH15 harbors the K and H T-DNAs oriented in an inverted repeat about the right border. The P35S-GUS cassette of the K T-DNA is located between two directly repeated loxP sites. After CRE-mediated deletion of the P35S-GUS cassette in line KH15, the Kd T-DNA in the KH15d6 line is generated. The parental line KH9/2 contains two transgene loci: one locus harbors the K and H T-DNAs oriented in an inverted repeat about the right border and one locus harbors a single H T-DNA copy. P35S, cauliflower mosaic virus promoter; 3′ 35S, 35S terminator; Pnos, promoter of the nopaline synthase gene; 3′-nos, 3′-end of the nopaline synthase gene; 3′-ocs, 3′-end of the octopine synthase gene; US, 792 nucleotides of the 3′-end of the GUS coding sequence; GUS, β-glucuronidase gene; HPT, hygromycin phosphotransferase gene; NPTII, neomycin phosphotransferase II gene; L, loxP sequence recognition site of CRE recombinase; LB, left border; RB, right border. attB1 and attB2, Gateway recombination sequences.

For the construction of the plasmids pK2L610 (K) and pH610 (H), the reader is referred to De Buck et al. (1998). The K T-DNA carries the neomycin phosphotransferase II (NPTII) gene under the control of the nopaline synthase promoter (Pnos), whereas this promoter drives the hygromycin resistance gene in the H T-DNA (De Buck et al., 1998). Both T-DNAs harbored the GUS gene under the control of P35S, but, in the K T-DNA, this P35S-GUS cassette was flanked by two directly oriented loxP sequences. Both T-DNAs had their borders in the natural octopine context.

Plant material – overview of the target lines

The pHhpUS T-DNA vector was transformed into the following transgenic Arabidopsis thaliana (L.) Heynh. lines: a transgenic ecotype Columbia (Col0) line carrying a P35S-GUS transgene with a Km-selectable marker at a single locus in the homozygous condition (FK24 line; Fig. 1) (De Buck et al., 2004); a transgenic C24 line carrying a CycB1;1-GUS transgene with a Km-selectable marker at a single locus in the homozygous condition (Ferreira et al., 1994); a transgenic C24 line carrying a CycD4;1-GUS transgene with a Km-selectable marker at a single locus in the homozygous condition (De Veylder et al., 1999); a transgenic C24 line carrying a KRP-GUS transgene with a Km-selectable marker at a single locus in the homozygous condition (L. De Veylder, VIB, Ghent University, Ghent, Belgium pers. comm.); a segregating seed stock of a C24 transgenic line for a WRKY23-GUS with a Km-selectable marker (Van de Cappelle et al., 2008).

Both parental lines KH15 and KH9/2 were identified after Arabidopsis C24 root transformation with two Agrobacterium strains carrying the K and H T-DNAs (Fig. 1) and subsequent selection on both kanamycin and hygromycin (De Buck et al., 2001; data not shown). In line KH15, both K and H T-DNAs were cointegrated at one genetic locus, but the H T-DNA was integrated in inverted orientation compared with the K T-DNA. In this manner, an inverted repeat of two T-DNAs around the right border was generated (De Buck et al., 2001; Fig. 1). Retransformation of this KH15 line with a C T-DNA, encoding the CRE recombinase, resulted in the KH15d6 transformant with a deletion of the P35S-GUS cassette in the K T-DNA (De Buck et al., 2001). PCR analysis revealed that this C T-DNA was absent in transformant KH15d6. Therefore, the deletion line KH15d6 was an isogenic form of locus KH15, but with one of the two inverted repeated P35S-GUS cassettes deleted (De Buck et al., 2001; Fig. 1).

Line KH9/2 contained two transgene loci: one, comparable with line KH15, harbored an inverted repeat of K and H T-DNAs around the right border, whereas the second locus had only one H T-DNA copy. Three different progeny plants were analyzed in more detail: KH9/2-1 containing only the inverted repeat locus; KH9/2-2 (described as line H9; De Buck et al., 2001) harboring a single H T-DNA copy; and KH9/2-3 containing both loci (Fig. 1).

Plant transformation, transformant selection and callus induction

The transgenic Arabidopsis lines FK24, C24:CycB1;1-GUS, C24:CycD4;1-GUS, C24:KRP-GUS and C24:WRKY23-GUS were supertransformed with the 35S-driven hairpin construct by the floral dip method (Clough & Bent, 1998). Seeds of the dipped plants were harvested and sown on K1 medium supplemented with hygromycin (20 mg l−1), resulting in the selection of 75 P35S-GUS/hpUS (P35S-hp), two CycB1;1-GUS/hpUS (CycB1;1-hp), 18 CycD4;1-GUS/hpUS (CycD4;1-hp), eight KRP4-GUS/hpUS (KRP4-hp) and five WRKY23-GUS/hpUS (WRKY23-hp) primary supertransformants. These supertransformants were self-fertilized and T2 seeds were collected. All plants were grown under a 16-h light:8-h dark regime at 21°C.

Seeds of the parental lines and T2-segregating hpUS supertransformants were germinated on Km-containing medium to select for plants that contain the original reporter promoter-GUS T-DNA. To identify the progeny plants that also contain the hpUS T-DNA, leaf tissue from all Km-resistant supertransformants was assayed for the ability to form callus on medium supplemented with hygromycin.

Callus was induced by placing leaf pieces on M1 medium containing 1 × Murashige and Skoog (MS) salts, 1 × MS vitamin mixture, 3% sucrose, 0.5 g l−1 2-(N-morpholino)ethanesulfonic acid (MES), 0.1 mg l−1 naphthaleneacetic acid, 1 mg l−1 benzylaminopurine and 0.7% agar (De Neve et al., 1997). Every 4–5 wk, a piece of each callus was taken and placed on fresh M1 medium.

Histochemical β-glucuronidase assay

Four-wk-old seedlings were fixed in 90% cold acetone for 30 min with continuous shaking. The seedlings were washed three times with 0.1 m Na2HPO4/NaH2PO4 buffer (pH 7) and incubated overnight at 37°C in 0.1 m Na2HPO4/NaH2PO4 buffer (pH 7) containing 10 mM EDTA, 0.5 mm K3[Fe(CN)6], 0.5 mm K4[Fe(CN)6] and 1% dimethylsulfoxide containing 50 mg ml−1 5-bromo-4-chloro-3-indolyl-β-d-glucuronic acid (X-gluc). Afterwards, the seedlings were washed with 0.1 m Na2HPO4/NaH2PO4 buffer (pH 7), destained in 90% ethanol and stored in 70% ethanol. Photographs were taken with a digital camera (AxioCam HRc; Zeiss, Jena, Germany) connected to a Zeiss Stemi SV11 microscope. To determine the percentage of root tips displaying no GUS staining, 2-wk-old seedlings were analyzed histochemically as described above. All seedlings were incubated for 15 min in X-gluc staining buffer.

To investigate and compare the effectiveness of the hairpin construct on the GUS activity in the root tips of the nine P35S-hp transformants, the numbers of GUS-negative and GUS-positive root tips were counted in 10 seedlings of the nine independent transgenic events. Data were analyzed by a generalized linear model with a binomial distribution and a probit link function. Except for P35S-hp51, P35S-hp61 and P35S-hp63, differences among the transformants were statistically significant (generalized linear model; χ2 test; < 0.001).

For detailed microscopy of GUS expression in root cells, 4-d-old seedlings were analyzed histochemically as described previously (Beeckman & Engler, 1994). All root samples were incubated overnight in X-gluc staining buffer, except for the roots of FK24 and P35S-hp seedlings, which were incubated for only 4 h in the staining buffer.

For median longitudinal sections, GUS-stained samples were treated as described previously (Beeckman & Viane, 2000), except that slides with sections were immersed in 0.05% ruthenium red for 10 min instead of toluidine blue O. Micrographs were taken with an Olympus BX51 microscope on a Nikon camera.

Preparation of protein extracts and determination of the GUS levels

Ground leaf and callus material was resuspended in 100 μl of buffer containing 50 mm phosphate buffer (pH 7), 10 mmβ-mercaptoethanol, 10 mm Na2-EDTA and 0.1% Triton X-100, and centrifuged twice at 4°C for 10 min to remove insoluble material. The total amount of soluble protein in the protein extracts was determined using the Bio-Rad Protein Assay (Bradford, 1976) with bovine serum albumin as a standard. The GUS activity was determined as described previously (Breyne et al., 1993). GUS activity levels were expressed as units of GUS protein relative to the total amount of soluble extracted protein (U GUS mg−1 protein).

Results

To determine the extent of GUS suppression induced by a hpRNA construct in different Arabidopsis cell types, we generated the 35S-driven hairpin construct of the last 792 nucleotides of the GUS coding sequence (the pHhpUS construct), and transformed it into five Arabidopsis parental lines already containing the GUS gene under the control of the 35S, CycB1;1, CycD4;1, KRP4 and WRKY23 promoters (Fig. 1; see Materials and Methods section). These promoters were chosen because they all differed in the cell types in which they promoted GUS expression (Table 1; Fig. 2). The 35S promoter drove transcription in practically all plant tissues; the CycB1;1 promoter in actively dividing tissues, such as the root meristem (root tip and emerging lateral roots), the shoot meristem and young leaves (Ferreira et al., 1994); the CycD4,1 and KRP4 promoters in the root tips and anthers only (De Veylder et al., 1999; Burssens et al., 2000); and the WRKY23 promoter in numerous, but not all, cell types (Att0001 in Barthels et al., 1997; Eulgem et al., 2000; Grunewald et al., 2008). Using histochemical staining of roots, leaves and flowers, we evaluated the efficiency of GUS suppression in five T2 seedlings of nine independent P35S-GUS/hpUS (P35S-hp), two CycB1;1-GUS/hpUS (CycB1;1-hp), five CycD4;1-GUS/hpUS (CycD4;1-hp), five KRP4-GUS/hpUS (KRP4-hp) and five WRKY23-GUS/hpUS (WRKY23-hp) supertransformants (Table 1; Figs 2c,e,g,3; see Materials and Methods section).

Table 1.   Overview of β-glucuronidase (GUS) activity in Arabidopsis thaliana parental lines and the hairpin supertransformants
ConstructRootLeavesFlower
  1. 1The most frequently occurring patterns are given, but note that there is variation among different supertransformants. For more details, see text.

In the parental lines
 P35S-GUSThroughoutAllThroughout
 CycB1;1-GUSEpidermal cells of the root tip and emerging lateral rootsYoungOvules and anthers (weak)
 CycD4;1-GUSColumella cellsNo expressionAnthers
 KRP-GUSColumella cellsNo expressionAnthers
 Lateral root cap cells  
 WRKY-GUSMain and lateral roots, including root tipsYoung; vascular tissue of expanded leavesThroughout the flower, especially in anthers and carpels
hp supertransformants1
 P35S-hpFrom no expression to throughoutNo expression or weak and patchyIn the style of the pistil
 CycB1;1-hpEpidermal cells of the root tip and emerging lateral rootsNo expressionMinor activity in the ovules and anthers (weak)
 CycD4;1-hpNo expression or only in the oldest columella cells or in the complete root tipNo expressionIn the style of the pistil
 KRP-hpNo expression or in columella cells and lateral root cap cellsNo expressionIn the style of the pistil
 WRKY-hpNo expression or main and lateral roots, including root tipsNo expression in older leaves, residual expression in young leavesThroughout the flower or only in the style of the pistil
Figure 2.

 Histochemical analysis of the parental lines and hairpin supertransformants. β-Glucuronidase (GUS) gene expression was visualized in the parental P35S, CycB1;1, CycD4;1, KRP4 and WRKY Arabidopsis thaliana lines (a, b, d, f) and in nine P35S-hp, two CycB1;1-hp, five CycD4;1-hp, five KRP4-hp and five independent WRKY23-hp supertransformants (c, e, g). Examples of GUS staining patterns in the parental lines and supertransformants are shown. For more details, see text.

Figure 3.

 Detailed β-glucuronidase gene (GUS) expression analysis in Arabidopsis thaliana root cells. GUS expression was visualized in the roots of 4-d-old parental P35S (a), CycB1;1 (b), CycD4;1 (c), KRP4 (d) and WRKY (e) parental lines, and in two P35S-hp (f, k), CycB1;1-hp (g, l), CycD4;1-hp (h, m), KRP4-hp (i, n) and WRKY23-hp (j, o) supertransformants. C, cortex; CRC, columella root cap; E, endodermis; Ep, epidermis; LRC, lateral root cap; PRC, peripheral root cap; Q, quiescent center; V, vascular tissue.

The hairpin construct suppresses GUS activity with different efficiencies in various plant tissues and with the lowest efficiency in root tips

For the parental lines, the histochemical results were as expected for the five promoters (Table 1; Figs 2a,b,d,f,3a–e). In the parental line FK24 (Fig. 1; see Materials and Methods section), the 35S promoter displayed GUS activity in all plant organs, but staining was especially intense in the root tips, rosette leaves and flowers. CycB1;1 exhibited GUS expression in the epidermal cells of the root tip and emerging lateral roots, in young leaves, in the ovules and only very weakly in the anthers of Arabidopsis. CycD4;1 displayed GUS expression in the columella cells of the root tips and in the anthers, whereas KRP4 induced expression in the columella and lateral root cap cells and anthers. Finally, WRKY23 exhibited GUS expression in root cells, including the root tips, in young leaves, in the vascular tissue of expanded leaves and in most floral tissues, especially in the anthers and carpels (Table 1; Figs 2a,b,d,f,3a–e).

The hairpin construct suppressed the GUS gene with widely varying efficiencies in different cell types (Table 1; Figs 2c,e,g,3f,o). In three P35S-hp supertransformants (P35S-hp/52, P35S-hp/55, and P35S-hp/57), GUS staining was absent to very low in the expanded roots and, on average, in more than 60% of the root tips (Fig. 4). Its intensity was strongly reduced in the vascular tissue and the columella cells of the root tips, but remained in the root epidermal cells (Figs 2c,3e). Almost all expanded leaves were GUS negative or showed a patchy pattern of weak GUS staining, which may indicate some cell-to-cell variability in the degree of silencing (Fig. 2e). In five P35S-hp supertransformants (P35S-hp/28, P35S-hp/51, P35S-hp/59, P35S-hp/61 and P35S-hp/63), GUS staining was observed in expanding roots and, on average, in 5–35% of the root tips (Fig. 4). This GUS staining was located in all root cells, except for the columella cells (Fig. 3k). Again no or weak and patchy GUS staining was visible in the leaves (Fig. 2e). The last supertransformant (P35S-hp/60) displayed GUS staining intensities in the roots similar to those in the parental FK24 line and in all root tips (Figs 2c,4), as well as a mixture of GUS-negative expanded leaves, patchy stained leaves and uniformly stained leaves (Table 1; Fig. 2e). In all nine P35S-hp supertransformants, GUS expression was suppressed in flowers, although residual GUS activity was always observed in the style of the pistil (Table 1; Fig. 2g).

Figure 4.

 Percentages of root tips displaying no β-glucuronidase (GUS) staining in 10 Arabidopsis thaliana T2 seedlings of nine P35S-hp supertransformants. GUS expression was visualized in 2-wk-old seedlings and the numbers of GUS-positive and GUS-negative root tips were calculated. In the parental FK24 line carrying the P35S-GUS transgene, 100% of the root tips were GUS positive. The graph shows the values (dots) for 10 T2 seedlings with the average (line). For supertransformants hp52 and hp61, only seven and eight seedlings, respectively, were analyzed. For each supertransformant, the numbers of seedlings with 100% and 0% white root tips are indicated.

In both CycB1;1-hp supertransformants CycB1;1-hp/1 and CycB1;1-hp/2, the GUS staining intensity was unaffected or slightly decreased in the roots, yet significantly decreased in the shoot meristem, in the leaves and in the ovules (Table 1; Figs 2c,e,f,3g,l). Of the five CycD4;1-hp supertransformants, three (CycD4;1-hp/1, CycD4;1-hp/3 and CycD4;1-hp/5) harbored root tips that were GUS negative or that displayed some remaining GUS activity in the oldest columella cells (Fig. 3h). The remaining two supertransformants (CycD4;1-hp/6 and CycD4;1-hp/10) showed all GUS-positive root tips, with GUS staining intensity in the columella root cells equal to that in the parental line (Fig. 3m). Similarly, plants of two KRP4-hp supertransformants (KRP4-hp/3 and KRP4-hp/4) showed roots without GUS staining, whereas for three supertransformants (KRP4-hp/1, KRP4-hp/2 and KRP4-hp/5), the GUS staining pattern did not change (Table 1; Figs 2c,3i,n). All CycD4;1-hp and KRP4-hp supertransformants maintained the same GUS staining in the style of the pistil (Fig. 2g). As both the CycD4;1 and KRP4 promoters are not active in leaves, the effect of hpRNA-induced silencing in leaves could not be evaluated (Table 1; Fig. 2d,e). In all WRKY23-hp supertransformants, GUS staining disappeared in older leaves, but some residual GUS staining was observed in young leaves (Table 1; Fig. 2e). In three supertransformants (WRKY23-hp/3, WRKY23-hp/5 and WRKY23-hp/6), GUS expression was suppressed in almost all root cells, except for the root tips, and in the style of the pistil (Figs 2g,3j,o). Indeed, detailed analysis of the root tips revealed the same GUS staining pattern in the quiescent center and the columella cells of WRKY-hp/5 as in the parental line, but total suppression in the vascular tissue (Fig. 3d). In addition, in all root tip cells, including the columella cells of supertransformant WRKY-hp/3, GUS activity was completely inactivated (Fig. 3j). In the other two supertransformants (WRKY23-hp/1 and WRKY23-hp/4), the GUS staining patterns in roots and flowers were comparable with those of the parental line (data not shown).

Quantitative evaluation of the hpRNA-mediated GUS suppression in leaf and callus tissues in the parental line and different P35S-GUS/hpUS supertransformants

To further investigate the efficiency of a 35S-driven hairpin GUS silencing construct in different cell types, GUS activity levels were determined in the leaves and calli of the parental Arabidopsis line FK24 and in nine independent P35S-hp supertransformants (as mentioned in the previous section). In line FK24, GUS activity was high, in both the leaves [mean, 1644 U GUS mg−1 total soluble protein (TSP)] and callus tissue (2557 U GUS mg−1 TSP) (Table 2; Fig. 5). In all nine supertransformants, except P35S-hp/60, GUS activity was low (<31 U GUS mg−1 TSP) and reduced by more than 120-fold in all T2 progenies compared with the GUS activity in the leaves of the parental line (Table 2; Fig. 5). In 50% of the T2 progeny plants of supertransformant P35S-hp/60, the GUS activity was reduced by only 25-fold (up to 80 U GUS mg−1 TSP), whereas, in the other 50%, the GUS activity was reduced by more than 100-fold (<10 U GUS mg−1 TSP) (Fig. 5). We can conclude that the 35S-hpUS construct generally induced efficient silencing of the 35S-driven GUS gene in expanding and mature leaves, confirming the histochemical staining results, and that, in the different supertransformants, the silencing strength imposed by the hpRNA in leaves correlated with the silencing efficiency observed in the roots.

Table 2.   β-Glucuronidase (GUS) activity in Arabidopsis thaliana leaves and calli of P35S-GUS transformants and P35S-hp supertransformants
Transgenic line1Leaves2 (4 wk old)Fold reduction3Calli2 (4 wk old)Fold reduction3
  1. GUS activity is expressed as units of GUS relative to the total amount of soluble extracted protein (U GUS mg−1 TSP).

  2. 1The transformants are ordered according to decreasing mean percentage of root tips without GUS staining (Fig. 4).

  3. 2Mean ± SD GUS activity; = 10.

  4. 3The ‘fold reduction’ is determined by the ratio of the mean GUS activity in the sample of the parental plant (P35S-GUS) to that in the sample of the supertransformant (P35S-hp).

  5. 4Bd, below detection, meaning below 0.1 U GUS mg−1 TSP.

P35S-GUS1644 ± 4052557 ± 239
P35S-hp/57 Bd4>1644010.2 ± 9.37251
P35S-hp/55 3.3 ± 9.849814.3 ± 12.4179
P35S-hp/52 Bd4>164402.93 ± 4.88872
P35S-hp/61 0.3 ± 0.85548019.8 ± 25.94130
P35S-hp/28 1.4 ± 2.5117447.6 ± 1454
P35S-hp/59 1.16 ± 2.311417135.5 ± 3919
P35S-hp/51 0.58 ± 1.28283469.1 ± 45.537
P35S-hp/63 13.58 ± 7.4121131.8 ± 130.3519
P35S-hp/60 27 ± 31.96194.4 ± 53.427
Figure 5.

 Quantitative β-glucuronidase (GUS) activity analysis in Arabidopsis thaliana leaves and calli of P35S-hp supertransformants. For the nine P35S-hp supertransformants, the hpUS construct efficiently suppressed the GUS gene in expanded leaves, but less uniformly in calli. GUS expression was measured in leaves (L) of 4-wk-old seedlings and in 4-wk-old calli (C). GUS activity levels are expressed as units of GUS per milligram of total soluble protein (U GUS mg−1 protein). The graph shows the values (dots) for 10 T2 seedlings with the average (line), and the transformants are ordered according to the average percentage of root tips without GUS staining (see Fig. 4).

The GUS activity in calli derived from the leaves of T2 plants of supertransformants P35S-hp/52, P35S-hp/55 and P35S-hp/57 remained low (<31 U GUS mg−1 TSP; Fig. 5) and was 179- to 872-fold lower than that in FK24 calli without a hairpin construct (Table 2). In calli of the supertransformants P35S-hp/28, P35S-hp/51 and P35S-hp/61, the GUS activity varied from undetectable to intermediate (<156 U GUS mg−1 TSP; Table 2; Fig. 5) among the induced calli, resulting in a mean reduction of 37- to 130-fold compared with the FK24 calli without a hairpin construct. Finally, in the calli of the remaining three supertransformants P35S-hp/59, P35S-hp/60 and P35S-hp/63, GUS activities were intermediate or high, and were only 19- to 27-fold reduced when compared with the calli of FK24. For instance, in all analyzed calli of P35S-hp/59 and P35S-hp/60, GUS activity levels varied between 80 and 220 U GUS mg−1 TSP and 26 and 185 U GUS mg−1 TSP, respectively (Fig. 5). In eight induced calli of supertransformant P35S-hp/63, GUS activity levels were between 26 and 374 U GUS mg−1 TSP. For the remaining two calli, no detectable GUS activity was observed (Fig. 5). Strikingly, the silencing strength imposed by hpRNA in the calli of different supertransformants ranked in a manner similar to that of the silencing efficiency in roots and leaves.

Quantitative evaluation of GUS suppression in leaf and callus tissues by inverted repeated GUS transgenes

In parallel, we examined the stability of transgene silencing induced by inverted repeated sense transgenes in Arabidopsis leaves and in profilerating callus cells during long-term propagation. Therefore, GUS expression was analyzed on leaves and calli of the homozygous and hemizygous parental line KH15, the homozygous and hemizygous deletion line KH15d6, the KH15×KH15d6 hybrids and the progeny plants of the parental line KH9/2 (Fig. 1) (De Buck & Depicker, 2001; De Buck et al., 2001) (see Materials and Methods section). In the homozygous and hemizygous deletion line KH15d6, harboring only one sense GUS transgene, GUS activity was high, in both leaves (> 1100 U GUS mg−1 TSP) and callus (> 5000 U GUS mg−1 TSP) (Table 3). In the hemizygous KH15 seedlings, harboring two inverted oriented GUS sense transgenes, the mean GUS expression in the leaves was reduced 40-fold compared with that in the KH15d6He plants, and was low (34 U GUS mg−1 TSP) throughout development (Table 3; data not shown) (De Buck et al., 2001). Furthermore, in the homozygous KH15 seedlings, the mean GUS expression in the leaves was reduced by more than 440-fold compared with that in the KH15d6He plants, and was very low (mean, 3 U GUS mg−1 TSP) throughout development (Table 3; data not shown) (De Buck et al., 2001), indicating a negative dosage effect of the inverted repeated transgene locus. On callus induction, GUS activity was eight- and two-fold lower in KH15Ho and KH15He, respectively, than that in transgenic KH15d6He calli containing only one sense transgene (Table 3). In addition, as in leaves, in the majority of calli, the GUS activity reached higher levels in the hemizygous KH15 calli than in the homozygous ones, confirming the negative dosage effect of the zygosity of the locus. Finally, the in trans silencing activity of the KH15 locus was high in KH15×KH15d6 leaves, because the GUS activity was reduced 266-fold when compared with that in KH15d6He leaves (Table 3). In KH15×KH15d6 induced calli, however, GUS activity was higher than in leaves, but still remained 11.8-fold lower than that in the KH15d6He calli, hinting at weak in trans silencing in calli (Table 3). When plants were regenerated from these calli, the increased GUS expression was not stably transmitted, indicating that the high GUS activity in callus was related to the reduced silencing efficiency in these tissues (data not shown).

Table 3.   β-Glucuronidase (GUS) activity in leaves and calli of Arabidopsis thaliana plants harboring inverted repeated sense transgenes or a single-copy sense transgene
TransgeneLeaves 4 wk old1Fold reduction2Calli 4 wk old1Fold reduction2
  1. GUS activity is expressed as units of GUS relative to the total amount of soluble extracted protein (U GUS mg−1 TSP).

  2. The number of analyzed seedlings is given in parentheses.

  3. 1, Mean ± SD.

  4. 2, The ‘fold reduction’ is determined by the ratio of the mean GUS activity in sample KH15d6 He to that in the corresponding sample.

  5. 3, Mean is within the experimental variation of the reference (De Buck et al., 2004).

KH15d6 He (15)1329 ± 4706343 ± 1925
KH15d6 Ho (15)1180 ± 761135015 ± 209113
KH15 Ho (15)3 ± 1443765 ± 7018
KH15 He (6)34 ± 39403305 ± 14371.9
KH15×KH15d6 (22)5 ± 8266534 ± 33011.8
KH9/2-2 (10)773 ± 6842854 ± 1171
KH9/2-1 (10)0.1 ± 07730199 ± 9314
KH9/2-3 (10) 0.1 ± 07730242 ± 18212

The same tendencies were seen for the progenies of the parental line KH9/2, containing two segregating loci (Table 3). GUS activity levels in 4-wk-old leaves of seedlings of KH9/2-1, containing an inverted repeat locus of two convergently transcribed GUS sense genes, and of KH9/2-3, containing the same inverted repeat locus and a single-copy locus, were very low (mean, < 0.1 U GUS mg−1 TSP), whereas the mean GUS activity levels in the leaves of KH9/2-2 seedlings, containing only the single-copy locus, were more than 7700-fold higher (773 U GUS mg−1 TSP) (Table 3). This observation implies that, in KH9/2-3 leaves, the inverted repeat locus is able to silence in trans the single H T-DNA copy. GUS activity in KH9/2-2 calli was 12- to 14-fold higher than that in the calli induced from KH9/2-1 and KH9/2-3 leaves, indicating that, also in these transformants, callus induction resulted in a significant, but partial, release of (in trans) silencing (Table 3).

Discussion

Constructs designed to produce dsRNA or self-complementary hpRNA transcripts efficiently induce targeted gene silencing (Helliwell et al., 2002; Stoutjesdijk et al., 2002). We investigated the efficiency of downregulation of a GUS gene expressed in different tissues of Arabidopsis by supertransforming a hairpin construct, containing the last 792 nucleotides of the GUS coding sequence, in several parental promoter-GUS lines. To study whether hpRNA-mediated gene suppression is equally efficient on different target GUS genes, an alternative approach could have been to cross all parental promoter-GUS lines with various types of promoters to the same 35S-hpRNA line. Although the variation between the different transformants would have been reduced, we chose not to apply it, because we wanted several 35S-hpRNA loci in combination with their target. Indeed, when the 35S-driven GUS gene was used as a target, GUS activity levels decreased significantly in all hpUS supertransformants, but the efficiency of GUS suppression varied in different supertransformants. Phenotypic series in independently obtained hairpin-containing Arabidopsis transformants have been stated to result from different T-DNA copy numbers, positional effects or effective interference initiated at different stages of plant development (Hilson et al., 2004). Indeed, multiple hairpin copies might be subjected to some degree of TGS, reducing their effectiveness in downregulating the homologous target genes (Kerschen et al., 2004; Small, 2007). As all the P35S-hp supertransformants harbored multiple hpUS T-DNA copies in one or two loci (data not shown), the hairpin locus structure might indeed explain the variable hairpin-mediated GUS silencing in the different supertransformants. The efficiency of silencing of a particular hairpin locus was stable and reproducible and the same ranking was found in leaves, roots and callus. In the P35S-hp supertransformants with no or very rare GUS staining in roots, GUS activity levels were efficiently suppressed in proliferating leaves, resulting in a 498- to more than 10 000-fold reduction, and in callus cells resulting in a 19- to 872-fold reduction, when compared with the GUS activity measured in the leaves and callus cells of the parental line FK24, respectively. However, in the P35S-hp supertransformant with GUS staining in all root tips, the silencing efficiency in leaves was 61-fold and in calli 27-fold higher than that in the parental line.

When the target GUS gene was driven by the CycB1;1, CycD4;1 and KRP4 promoters, hpUS-mediated GUS suppression was only limited and almost never complete in the root tips. In addition, GUS suppression of the WRKY23-GUS target gene was efficient in expanding root tissues, but not or weakly in root tips, young leaves and flowers. Currently, we cannot explain why the 35S-driven GUS gene is more efficiently suppressed than cell cycle-specific, promoter-driven GUS genes, in particular because the cell cycle promoters CycB1;1, CycD4;1 and KRP4 are only weakly active in the root tips when compared with the P35S promoter (L. De Veylder, pers. comm.). Thus, GUS mRNAs are expected to be less abundant in these than in the 35S-GUS transformants. Root tips with no detectable GUS activity were found in some P35S-hp, CycD4;1-hp, KRP4-hp and WRKY23-hp supertransformants, implying that all cell types are susceptible to transgene silencing by the hpUS construct. In the majority of supertransformants, however, GUS expression was clearly suppressed in mature leaf and root tissues, but not or only partially suppressed in the root tips. This observation suggests that transcripts in root tip-located cells are the most resistant to hairpin-mediated silencing. In addition, in single T-DNA copy Arabidopsis plants exhibiting transgene silencing on the basis of dosage effects, the time point of downregulation differed for leaves and roots, and downregulation was later and much less pronounced in roots than in leaves (De Wilde et al., 2001). This observation indicates that silencing on the basis of RNA threshold levels occurs less efficiently in roots than in leaves (De Wilde et al., 2001). Moreover, the RNA silencing-mediated resistance of Nicotiana benthamiana to beet necrotic yellow vein virus seemed less effective in roots than in leaves, because higher levels of transgene mRNAs and lower levels of transgene-derived siRNAs accumulated in roots than in leaves of silenced plants (Andika et al., 2005). This lower transgene RNA silencing activity in roots was also associated with lower transgene methylation levels at a nonsymmetrical CpNpN context, suggesting that roots are less active in the generation of siRNAs (Andika et al., 2006).

When looking at the GUS staining patterns in the root tips in the CycD4-hp, KRP4-hp and WRKY-hp supertransformants in which the GUS gene was not efficiently suppressed, we found it especially striking that the columella cells in the hairpin supertransformants still showed an almost equal staining intensity to that of the parental plants, whereas silencing occurred efficiently in the vascular tissue cells, suggesting that RNA in columella cells might be less sensitive to in trans RNA silencing. Dunoyer et al. (2006) postulated that phytohormones, including auxins, might be essential for silencing suppression in tumors induced by Agrobacterium tumefaciens infection. In roots, an auxin gradient, with a maximum in the Arabidopsis root apex, could be demonstrated (Petersson et al., 2009). The highest relative auxin concentration was observed in the quiescent center, with high levels in the cortex, endodermis and apical parts of the stele, whereas, in the columella and epidermis, auxin levels were lower than those on average in the surrounding tissues. Therefore, it is tempting to speculate that the low concentration of auxin in the columella cells makes them less sensitive to RNA silencing. Nevertheless, the opposite was true for the P35S-hp supertransformants with residing GUS expression in the root tips. Here, all root tips showed efficient silencing in the columella cells, whereas staining was still present in all remaining cells of the root tips in the supertransformants. At present, this discrepancy in inactivation patterns in the root tip cells cannot be explained. Possibly, the promoter homology between the 35S-GUS construct in the target plant and the hpUS construct might cause transcriptional silencing. In the CycD4-hp, KRP4-hp and WRKY-hp supertransformants, only homology in the GUS coding sequence was present, hinting at PTGS only. Kerschen et al. (2004) suggested that factors, such as spatial and temporal gene expression patterns, might affect RNAi effectiveness. Indeed, Nakatsuka et al. (2007) produced various flower pigmentation patterns using a specific promoter and, in this manner, controlled the tissue specificity of dsRNA formation. Here, the target RNA is in each case the GUS mRNA, but produced by different promoters. Thus, we might hypothesize that the susceptibility of the target RNA is codetermined by the promoter that specifies the timing and abundance profile, possibly even the localization of the mRNA. Although one could argue that the not truly constitutive character of the 35S promoter codetermines the variable silencing efficiencies, we are convinced that the different cell types themselves are responsible for this variability, because both the FK24 parental line and the hairpin construct are driven by the same 35S promoter. In addition, for all target promoters, the lack of inactivation is especially prominent in cell types of the root tips and flowers, whereas it is clear that the 35S promoter is fully active in these cells.

A drastic release of in cis and in trans gene silencing efficiency was observed in the calli of transformants harboring two inversely oriented T-DNAs with convergently transcribed GUS transgenes. Although the GUS silencing efficiency in leaves was between 40- and 7000-fold in all analyzed cases, the silencing efficiency in the respectively derived calli was only c. two- to 14-fold, demonstrating that the silencing degree of inverted repeated sense transgenes in leaves is much stronger than in derived calli. In addition, in transgenic rice (Oryza sativa) calli, hairpin constructs induced much more effective silencing than simple sense and antisense constructs (Wang & Waterhouse, 2000). The different silencing efficiencies induced by hairpin constructs and inverted repeated sense transgenes in calli might be explained by the manner in which they produce dsRNAs. Hairpin constructs directly generate dsRNAs, whereas inverted repeated sense transgenes require RNA-dependent RNA polymerase for the generation of dsRNAs (Dalmay et al., 2000; Mourrain et al., 2000; Butaye et al., 2004; Horiguchi, 2004; Bleys et al., 2006b). Thus, we speculate that one or more elements or processes contributing to the efficiency of silencing might be impaired in proliferating callus cells, such as the production of dsRNA, RNA-induced silencing complex-mediated RNA degradation and/or repression of translation. Dunoyer et al. (2006) also postulated that siRNA synthesis is specifically inhibited in tumors, as Dicer activities producing siRNAs from inverted repeat transcripts were specifically suppressed in these tumors. Another explanation for the different silencing efficiencies induced by hairpin constructs and inverted repeated sense transgenes might be the difference in structure between the silencing constructs. Although in the hpRNA construct the intron has been spliced out, resulting in a hairpin sequence without spacer region between the palindromic sequences, a bulge of 431 bp was present in the inverted repeated T-DNAs as a result of the presence of an additional 3′-end region in the K T-DNA. Indeed, enhanced hpRNA-induced silencing has already been reported to be obtained using an intron instead of a GUS or green fluorescent protein spacer (Smith et al., 2000; Wesley et al., 2001). In addition, an internal rearrangement of an Arabidopsis inverted repeated phosphoribosylanthranilate isomerase (PAI) locus led to suppressed transcription of the inverted repeated locus and reduced methylation of PAI-identical sequences (Melquist & Bender, 2004). Furthermore, we have suggested previously that a spacer in between inverted repeated transgenes strongly reduces the degree and stability of silencing (De Buck et al., 2001), although the frequency of RNA silencing has been found to be affected more strongly by the spacer sequence than by its spacer length; inverted repeats interrupted by a spacer region up to 1022 bp gave efficient silencing (Chuang & Meyerowitz, 2000; Hutvágner et al., 2000; Hirai et al., 2007).

In summary, the 35S-driven hairpin construct can produce a series of independent Arabidopsis transformants with different degrees of silencing. In some hpUS supertransformants, the 35S-GUS target gene was efficiently suppressed in expanding leaf and root tissues, but less in shoot and root tips. When the hpUS construct was targeted against the GUS gene driven by the CycB1;1, CycD4;1 and KRP4 promoters, hpUS-mediated GUS suppression was only limited, and, in particular, the GUS activity in root tips could not be downregulated efficiently. Finally, the imposed silencing of inverted repeated sense genes was as strong as that of an hpRNA construct in leaves, but weaker in callus cells.

Acknowledgements

The authors thank Marieke Louwers and Anni Jacobs for practical assistance, Professor Marnik Vuylsteke for help with the statistical analysis, Dr Annick Bleys and Leen Vermeersch for critical reading of the manuscript, Karel Spruyt for photographic work, Jean-Luc Doumont for editing and Dr Martine De Cock for help in preparing the manuscript. This work was supported by grants from the European Union BIOTECH program (QLRT-2000-00078) with additional cofinancing from the Flemish Community, the ‘Bijzondere Onderzoeksfonds’ of Ghent University (BOF 01111400) and the Research Foundation-Flanders (no. G.021106).

Ancillary