Spreading of post-transcriptional gene silencing along the target gene promotes systemic silencing

Authors

  • Rubén Darío García-Pérez,

    Search for more papers by this author
      Present address: National Institute for Forestry, Agriculture and Livestock Research (INIFAD), Carretera Culiacán-El Dorado km 16.5, Culiacán, Sinaloa, México.
  • Helena Van Houdt,

    1. Department of Plant Systems Biology, Flanders Interuniversity Institute for Biotechnology (VIB), Ghent University, Technologiepark 927, B-9052 Gent, Belgium
    Search for more papers by this author
  • Anna Depicker

    Corresponding author
      For correspondence (fax +32 9 331 3809; e-mail ann.depicker@psb.ugent.be).
    Search for more papers by this author

For correspondence (fax +32 9 331 3809; e-mail ann.depicker@psb.ugent.be).

Present address: National Institute for Forestry, Agriculture and Livestock Research (INIFAD), Carretera Culiacán-El Dorado km 16.5, Culiacán, Sinaloa, México.

Summary

Transitive silencing and grafting-induced gene silencing phenomena were combined to investigate whether a primary target β-glucuronidase (gus) gene could promote the generation of systemic transitive silencing signals. Tobacco plants with hemizygous or homozygous silencer locus and in trans silenced primary target were used as a source of post-transcriptionally silenced rootstocks and tobacco plants with or without a secondary target locus as scion source. The silencer locus harbored two identical neomycin phosphotransferase II (nptII)-containing T-DNAs, integrated as an inverted repeat. The primary target locus carried a gus gene with homology to the transcribed region of the nptII gene only in the 3′ untranslated region, whereas the secondary target locus had two or more copies of a gus gene without homology to transcribed sequences of the silencer locus. The upstream region of the initially targeted sequences of the in trans silenced gus gene could induce the production of a systemic signal. This signal was capable of triggering post-transcriptional gene silencing (PTGS) of the secondary target gus genes in the scion. In addition, the induction of systemic silencing was strikingly dosage dependent for the silencer as well as the primary target loci in the rootstock. Moreover, in the scions, the secondary target gus genes had to be present to generate detectable amounts of short interfering RNAs.

Introduction

Post-transcriptional gene silencing (PTGS) is a sequence-specific degradation process of RNA derived from endogenes and/or transgenes in plants transformed with homologous sequences of these genes. Viruses can induce and become targets of a similar PTGS mechanism (Dougherty et al., 1994; Kumagai et al., 1995; Lindbo et al., 1993), in which double-stranded RNA (dsRNAs) plays a central role. Viruses form dsRNAs during replication, whereas plant transgenes may give rise to dsRNA by self-complementarity of single-stranded RNA transcripts, by hybridization of sense and antisense sequences, or by an RNA-dependent RNA polymerase (RdRP)-mediated synthesis that uses aberrant RNA as the template (Montgomery and Fire, 1998; Stam et al., 2000; Vaucheret et al., 2001; Waterhouse et al., 1998). In animals, an analogous RNA degradation mechanism, called RNA interference (RNAi), is also efficiently triggered by dsRNA. The dsRNA molecules are processed by an RNase III-like enzyme (designated DICER) into fragments of 21–23 nucleotides (Zamore et al., 2000), designated short interfering RNAs (siRNAs). Biochemical analysis showed that the siRNAs are incorporated into an RNA-induced silencing complex to guide degradation of complementary mRNAs (Hammond et al., 2000). As a result, such target RNAs are cut approximately in the middle of the region that is complementary to the siRNA (Elbashir et al., 2001). Small RNAs (smRNAs), previously found in plants undergoing PTGS induced by viruses and transgenes (Hamilton and Baulcombe 1999), have recently been shown to belong to two different size classes (Hamilton et al., 2002; Mallory et al., 2002). By using wheat germ extracts, Tang et al. (2003) found that different DICER-like enzymes convert dsRNAs into siRNAs of the two size classes.

Strikingly, a few dsRNA molecules are sufficient to induce RNAi in Caenorhabditis elegans, implying that they can act catalytically and/or that they are amplified (Fire, 1999; Fire et al., 1998). Indeed, the pool of siRNAs is amplified in nematodes that have been injected with particular dsRNAs. Non-homologous siRNAs are generated, which correspond specifically to the adjacent upstream region of the original target sequence, suggesting an amplification cycle in which an RdRP can use the siRNAs as primers and the target mRNAs as templates to produce dsRNAs. These newly formed dsRNAs are subsequently cleaved to produce secondary siRNAs (Lipardi et al., 2001; Sijen et al., 2001). Also, in some cases of virus-induced gene silencing (VIGS) in plants, downregulation of the target gene persists after the viral RNA has been eliminated (Ruiz et al., 1998) and correlates with spreading of RNA targeting along the transgene transcript (Vaistij et al., 2002). VIGS can spread to the 3′ and 5′ regions of the originally targeted sequence within the mRNA, implying primed and unprimed amplification cycles (Vaistij et al., 2002; Voinnet et al., 1998), which is in accordance with the observed in vitro activity of an RdRP isolated from tomato (Schiebel et al., 1993, 1998). Transgenes and siRNAs have also been reported to induce spreading of gene silencing along a transgenic target gene (Braunstein et al., 2002; Klahre et al., 2002; Van Houdt et al., 2003), although endogenous targets do not support spreading upon VIGS (Vaistij et al., 2002). However, secondary siRNAs may be generated on the basis of endogene-derived transcripts, as suggested by Sanders et al. (2002).

In specific well-described cases, PTGS can be initiated in a specific area of the plant and then spread via limited and extensive cell-to-cell movement (Himber et al., 2003), and via systemic transport (Palauqui et al., 1997; Voinnet et al., 1998). It has been hypothesized that the silencing signal might be amplified because silencing can persist a long time after the original source of silencing has been eliminated (Palauqui and Balzergue, 1999). Grafting and Agrobacterium infiltration experiments revealed a systemic silencing signal that is sequence specific, moves over long distances (Palauqui et al., 1997) bidirectionally through plasmodesmata and phloem (Voinnet et al., 1998), and is efficiently recognized when the target mRNA from endogenes or transgenes accumulates at high levels (Palauqui and Vaucheret, 1998). However, a systemic signal is seemingly not an essential feature of gene silencing mediated by antisense transgenes and sense amplicons (Crétéet al., 2001; Fagard and Vaucheret, 2000a; Mallory et al., 2003) and of transcriptional gene silencing (Fagard and Vaucheret, 2000b; Mlotshwa et al., 2002). Separate signal molecules have been proposed for cell-to-cell and long-distance systemic transport (Himber et al., 2003). Hamilton et al. (2002) suggested that long smRNAs (24–26 nucleotides) are part of the long-distance signal because their reduction by viral suppressors negatively affects systemic silencing. In contrast, the plant's capacity for systemic silencing has recently been documented not to correlate with the accumulation of long smRNAs (Mallory et al., 2003), leaving unanswered the question about the exact nature of the mobile signal.

Previously, we have demonstrated that the in trans silencing capacity of a silencer locus (X) can be transmitted to a primary target (Y) that can subsequently silence a secondary target (Z) in trans (Van Houdt et al., 2003). The gus genes of the secondary target locus Z, without transcribed sequences homologous to the silencer locus X, become silenced only when the primary target locus Y, which produces transcripts homologous to those of loci X and Z, is also included into the same plant genome. To study the spreading of PTGS along the target gene and over a long distance through the plant, we combined spreading of silencing along the primary target gene (transitive silencing) with the grafting-induced silencing phenomena. In particular, we investigated whether the primary target gene in locus Y could promote the generation of systemic transitive silencing signals.

Results

Experimental setup to study the systemic nature of transitive silencing

Previously, we have shown the occurrence of transitive silencing in transgenic tobacco plants carrying three transgene loci X, Y, and Z, which exhibit stepwise homology in the transcribed sequences of the encoded reporter genes (Van Houdt et al., 2003). The post-transcriptional silencer locus X does not only silence a primary target gus gene in locus Y because of homology in the 3′ region but also trigger spreading of the silencing-inducing capacity along the target RNA, with transitive silencing as a result. As such, the secondary target gus genes in locus Z, which have no transcribed regions homologous with the silencing-inducing gene in locus X, become efficiently silenced.

To determine whether an in trans silenced target Y1 (corresponding to Y; Van Houdt et al., 2003) could promote the second step of the transitive silencing in a systemic manner, we combined the three loci (X, Y1, and Z), necessary to assay for transitive silencing, by grafting a locus Z-containing scion (ZZ) onto a locus X- and Y1-containing rootstock.

The three loci X, Y1, and Z used to study PTGS spreading along the target gene and through the plant are represented in Figure 1. Locus X is a very efficient post-transcriptional silencer locus that carries a T-DNA inverted repeat toward the right border with two convergently transcribed P35-neomycin phosphotransferase II (nptII)-3′ chalcone synthase (chs) transgenes; locus Y1 harbors a single copy of a relatively highly expressed P35S-gus-3′chs transgene, whereas locus Z carries two or more copies of a moderately expressed P35S-gus-3′ nopaline synthase (nos) transgene. To make the grafts, tobacco plants of approximately 30 cm height with the silencer locus X and the in trans-silenced primary target Y1 in hemizygous or homozygous condition (XXY1Y1_, XXY1_, X_Y1Y1, and X_Y1_; the underscore symbolizing the hemizygous condition of the transgene locus) were used as source of gus-silenced rootstocks; plants approximately 10 cm tall homozygous for locus Z (ZZ) served as gus-expressing scions (Figure 2). At least 9 and up to 16 independent grafts were analyzed for each combination (Figure 3).

Figure 1.

Schematic outline of the T-DNA constructs (drawn to scale) present in the silencer locus X, primary target locus Y1, and secondary target locus Z (T-DNAs of pGVCHS287, pGUSchsS, and pXD610, respectively), as well as of the homology between the transcribed sequences of the P35S-driven genes of X, Y1, and Z.

The structure of loci X and Y1 is indicated; locus Z contains two or more copies of the XD610 T-DNA. 3′chs, 3′ untranslated region (UTR) of the chs gene of snapdragon (A. majus); Δ, 3′chs polyadenylation signal; 3′ g7, 3′ UTR of the Agrobacterium tumefaciens octopine T-DNA gene 7; 3′nos, 3′ UTR of the nos gene; 3′ocs, 3′ UTR of the octopine synthase gene; bar, bialaphos acetyltransferase-coding sequence conferring phosphinothricin resistance; gus c.s., gus-coding sequence; hpt, hygromycin phosphotransferase-coding sequence; I, artificial intron; LB, left T-DNA border; nptII c.s., neomycin phosphotransferase II-coding sequence; P35S, CaMV 35S promoter; Pnos, nopaline synthase promoter; Pss, promoter of the small subunit of ribulose-1,5-bisphosphate carboxylase; and RB, right T-DNA border.

Figure 2.

Schematic representation of the grafting procedure used.

(a) Plants of approximately 30 and 10 cm height, with the genotypes indicated, were used as source for rootstocks and scions, respectively. Scissors designate the place where the stem was cut to make the graft (see Experimental procedures).

(b) Graft junction, secured with parafilm, and genotype of the resulting grafts. /, grafted on top of.

Figure 3.

GUS activity levels in ZZ scions (a) and percentage of gus-silenced ZZ scions (b) 6 weeks after being grafted onto (/) rootstocks with or without a post-transcriptionally silenced gus gene. The four different types of gus-silenced rootstocks carried a silencer locus X and a primary target gus gene in locus Y1, both in homozygous or hemizygous conditions. In the non-gus-silenced rootstocks used as controls, locus X or Y1 were homozygous (XX and Y1Y1) or X and Y2 hemizygous (X_Y2_).

(a) SD is given by error bars.

(b) A plant was considered silenced when its GUS activity was lower than the mean of all the control ZZ scions minus twice the SD. N, total number of grafts tested in each case; n, number of silenced ZZ scions; and TSP, total soluble protein.

PTGS of a primary target gene induces production of systemic signals

The first aim of this study was to test whether the in trans-silenced primary target Y1 (Figure 1) was able to induce the production of systemic transitive silencing signals. Therefore, plants containing the silencer locus X and the primary target in locus Y1, both in homozygous condition (XXY1Y1), were used as source of gus-silenced rootstocks, whereas those carrying the secondary target in locus Z, in homozygous condition (ZZ), as source of gus-expressing scions, resulting in the grafts designated ZZ/XXY1Y1 (Figures 2 and 3). The ZZ scions, grafted onto XX, Y1Y1, or X_Y2_ (hemizygous) non-gus-silenced rootstocks functioned as controls (for locus Y2, see Experimental procedures).

To show the in trans inactivation of the gus genes in Y1 by the silencer X in the XXY1Y1 rootstocks, the GUS activity was determined in protein extracts from samples collected 3 weeks after grafting. As controls, Y1Y1 rootstocks were analyzed. The GUS activity in XXY1Y1 rootstocks was 1.1 ± 0.8 U GUS mg−1 of total soluble protein (TSP), which is much lower than that in Y1Y1 rootstocks with a mean value of 1344 ± 356 U GUS mg−1 TSP. Six weeks after grafting, the GUS activity in ZZ scions was determined. The mean GUS activity values in the ZZ scions grafted onto XX, Y1Y1, or X_Y2_ (used as controls), were high and statistically similar, as expected (Figure 3a). The criterion for discriminating silenced from non-silenced ZZ scions grafted on gus-silenced rootstocks was a GUS activity value below the mean of the pooled control samples minus twofold the standard deviation. On the basis of this criterion, almost 90% (14/16) of the ZZ scions grafted onto XXY1Y1 rootstocks were silenced (Figure 3b). The mean GUS activity of the silenced ZZ scions was 7% of that of the controls (Figure 3a). This result clearly indicates the ability of a silenced target gene to produce systemic silencing signals.

Production of the systemic signal depends on the target dosage

Systemic silencing in plants has been observed by using grafting, Agrobacterium infiltration, and biolistics systems (reviewed by Mlotshwa et al., 2002). Nevertheless, the factors involved in the transition from local to systemic silencing are not well known. Considering that a target gene can induce the generation of systemic transitive silencing signals, we examined whether the production of the systemic signal could depend on the dose of X and Y1 targets. Therefore, we also analyzed grafts with X_Y1_, XXY1_, and X_Y1Y1 rootstocks as source of X and Y1 dosage variation, with ZZ scions as secondary targets. Reduction of the GUS activity from the Y1 locus in X_Y1_, XXY1_, and X_Y1Y1 rootstocks was very pronounced 3 weeks after grafting (2–5 U GUS mg−1 TSP), indicating that silencing in trans of X upon Y1 had been established as efficiently as in the previously mentioned XXY1Y1 rootstocks.

Six weeks after grafting, 89 and 100% of the ZZ scions grafted onto XXY1_ and X_Y1Y1 rootstocks showed GUS silencing, respectively, which is similar to the 88% silenced ZZ scions obtained upon grafting onto XXY1Y1 rootstocks (Figure 3b). In contrast, none of the ZZ scions grafted onto X_Y1_ rootstocks exhibited GUS silencing; their mean GUS activity was statistically similar to those of the control grafts (Figure 3a,b), whereas those of the silenced ZZ scions grafted onto XXY1_ and X_Y1Y1 rootstocks corresponded to that of the silenced ZZ scions grafted onto XXY1Y1 rootstocks. In conclusion, only a double dose of the silencer X or the primary target Y1 in the rootstock results in systemic transitive silencing in ZZ scions.

Production of secondary siRNAs depends on target dosage

Transgene- and endogene-derived mRNAs can, in some cases, become a production source for secondary siRNAs (Sanders et al., 2002; Sijen et al., 2001; Vaistij et al., 2002; Van Houdt et al., 2003). However, the specific factors involved in producing detectable amounts of secondary siRNAs from primary target RNAs are not known. To determine whether transgene dosage plays a role in the production of secondary siRNAs and whether the amount of siRNAs produced is correlated with the efficiency of systemic silencing, accumulation of siRNAs was evaluated in XXY1_, X_Y1_, XXY1Y1, and X_Y1Y1 rootstocks that were used as a source of transgene dosage variation, and also in an Y1Y1 rootstock used as a negative control.

As shown in Figure 4a, leaves of the XXY1Y1 and X_Y1Y1 rootstocks produced approximately twofold the amount of secondary gus siRNAs (lanes 3–5) compared with that of the XXY1_ and X_Y1_ rootstocks (lanes 1–2 and 6–7, respectively), illustrating that the accumulation level of secondary gus-specific siRNAs depends on the Y1 target dose. Besides, no gus-specific siRNAs were detected in the control rootstock Y1Y1 (Figure 4a, lane 8). However, the gus-specific secondary siRNA accumulation level in the XXY1_ rootstocks (which can send efficiently the systemic transitive signal) with the double dose of the silencer locus X was not significantly higher than in X_Y1_ rootstocks with the single dose X (Figure 4a, compare lanes 1 and 2 with lanes 6 and 7). Thus, there is no clear correlation between the presence or the amount of gus-specific siRNAs in the rootstock and the ability to induce systemic silencing.

Figure 4.

Accumulation of gus smRNAs.

(a) The relative mean value of gus smRNAs per genotype ± the SD is indicated by the arrowhead. The accumulation of gus smRNAs in rootstocks containing locus X plus Y1 was approximately twofold higher in rootstocks carrying Y1 in homozygous condition (lanes 3–5) than in those carrying Y1 in hemizygous condition (lanes 1, 2, 6, and 7). No gus smRNAs were detected in the control rootstock Y1Y1 (lane 8).

(b) Gus smRNAs were detected in ZZ scions grafted onto XXY1Y1 and X_Y1Y1 rootstocks (lanes 4 and 5), but neither in SR1 scions grafted onto X_Y1Y1 rootstocks (lanes 1 and 2) nor in the control ZZ scion grafted onto an Y1Y1 rootstock (lane 3). nt, nucleotides.

Low molecular weight RNA was extracted from leaves of rootstock (a) and scion (b) 7 weeks after grafting (see Experimental procedures). DNA oligomers of 21 and 25 nucleotides in length were used as molecular markers. smRNAs were detected with a hydrolyzed 32P-labeled sense riboprobe comprising the full-length gus-coding sequence (GusS probe). The predominant ethidium bromide (Etbr)-stained species in the low molecular weight RNA fraction are shown as loading controls.

The presence of the secondary target gene is required to generate detectable amounts of siRNAs in the scion

In case silencing is induced by a mobile signal, the siRNAs detected in systemically silenced tissues can originate from the rootstock or can be the product of amplification in the scion, with the target gene-derived RNAs and/or the systemic signal itself as template. To determine whether the systemic silencing signal, in the absence of a target gene, could induce the accumulation of detectable amounts of siRNAs, wild-type (SR1) scions were grafted onto X_Y1Y1 rootstocks. As positive and negative controls, ZZ scions were grafted onto XXY1Y1 or X_Y1Y1 rootstocks and onto Y1Y1 rootstocks, respectively.

Gus-specific siRNAs were detected in ZZ scions grafted onto XXY1Y1 or X_Y1Y1 rootstocks (Figure 4b, lanes 4–5), but not in wild-type SR1 scions grafted onto X_Y1Y1 rootstocks (two of the four tested are shown in Figure 4a, lanes 1 and 2), not even when 100 μg per lane instead of 25 μg per lane of low-molecular-weight RNA was loaded (data not shown). Together, taking the detection limit of the assay into account, in wild-type SR1 scions gus-specific siRNAs are at least 15-fold less abundant, if present at all. Gus-specific siRNAs were also absent in the negative control scion (Figure 4b, lane 3). From these observations, we deduced that most likely the gus transgenes of locus Z in the scion, which act as secondary targets, are required to amplify the systemic transitive signal from the rootstock into a pool of siRNAs in the scion. As such, the siRNAs formed in the scion can be considered as tertiary siRNAs.

Discussion

Signal amplification in the rootstock

By using the grafting technique, we demonstrate that rootstocks carrying a primary target gus gene in locus Y1 and a silencer locus X are capable of sending a transitive silencing signal to induce PTGS of gus in scions carrying locus Z. The identity of the signal is unknown, but most probably includes an RNA molecule that gives sequence specificity to the signal (reviewed by Mlotshwa et al., 2002). We postulate that the RNA component of the systemic signal could be produced by marking the primary target gus transcripts from locus Y1, possibly via base pairing with complementary 3′chs siRNAs derived from locus X (Van Houdt et al., 2003), and by using those marked transcripts as templates by an RdRP-like activity. In this way, the amplification reaction could give rise to secondary siRNAs, directly by making short complementary RNAs or indirectly via dicing of an extended dsRNA precursor (Makeyev and Bamford, 2002). Recently, it has been shown (Himber et al., 2003) and confirmed by our results that secondary siRNAs belong exclusively to the 21-nucleotide siRNA class; this observation argues against the previously proposed role for the 25-nucleotide siRNAs as systemic silencing signal (Hamilton et al., 2002; Himber et al., 2003). Together, a functional role is being implied for RNA amplification products, such as 21-nucleotide secondary siRNAs or precursor molecules, in systemic silencing.

We found that the silenced target promotes production of systemic signals, only when the silencer and/or the target locus are present in homozygous condition. With the amplification model described previously, these results could be explained if marking the target gus transcripts by the 3′chs siRNAs were a limiting factor to produce sufficient signal for systemic silencing. Thus, production of the systemic signal would require that gus transcripts generated by the primary target locus Y1 were marked by the 3′chs siRNAs produced by locus X and Y1. In this scenario, X_Y1Y1, XXY1_, and XXY1Y1 rootstocks would produce more gus transcripts, or 3′chs siRNAs, or both than the X_Y1_ rootstocks, so the probability of the gus transcripts and the complementary 3′chs siRNAs to find each other would be higher in the first three cases than in the last one. In this context, X_Y1_ would produce enough 3′chs siRNAs and gus-3′chs transcripts to amplify the pool of siRNAs used to trans-inactivate the gus gene in locus Y1, but not to produce enough of the suggested RNA component of the systemic silencing signal. This hypothesis is in concordance with the induction of local but not systemic silencing when a small amount of DNA is bombarded in plants that are capable of producing a systemic signal (Palauqui and Balzergue, 1999). Also, during transitive RNAi in C. elegans molecules are produced that can move between cells. Moreover, the severity of the phenotype obtained upon transitive RNAi of a particular endogene seems to depend on the abundance of the transgenic primary target mRNA (Alder et al., 2003).

Taken together, our results from tobacco grafting experiments and data obtained in C. elegans (Alder et al., 2003) indicate that a systemic silencing signal is not necessarily produced by the silencer locus, but that a target gene can promote the generation of a systemic signal and that a certain threshold of the target gene seems to be required to induce systemic silencing. This proposition could explain why spreading of grafting-induced silencing to newly developed leaves after re-grafting onto wild-type rootstocks (referred to as ‘maintenance’ by Palauqui and Vaucheret, 1998) occurred in homozygous, but not in hemizygous plants that can trigger silencing spontaneously at various frequencies in homozygous condition. Moreover, the reason why sense- and not antisense-mediated gene silencing produced systemic silencing (Crétéet al., 2001) may be that the target RNAs are highly expressed in the former and probably unstable in the latter.

Transgenic and wild-type plants can recover from virus infection and produce new virus-free leaves that are resistant to infection by an identical or a related virus. A systemic silencing signal has been suggested to go ahead of the infection front to prevent spreading of the virus (Voinnet et al., 2000). Taking this hypothesis into account, the model described above, in which amplification of the signal depends on the target dose, could clarify why transgenic plants recover more efficiently when they carry more copies of the transgene homologous to the infecting virus (Goodwin et al., 1996).

Signal amplification in the scion

As no detectable amounts of gus siRNAs were found in wild-type SR1 plants grafted onto silenced gus rootstocks able to send a systemic signal (Figure 4b, lines 1–2), we propose that the siRNAs observed in the ZZ scions grafted onto X_Y1Y1 or XXY1Y1 rootstocks (Figure 4b, lanes 4–5) are the products of amplification in the scion; therefore, they are designated as tertiary siRNAs. Thus, the secondary target gene-derived mRNAs may be required for amplification to establish gene silencing in response to the mobile signal. This hypothesis is in agreement with the fact that the target gene in the scion needs to be transcriptionally active (Palauqui and Vaucheret, 1998), and that viruses homologous to a silenced gene in the rootstock are only targeted for degradation when inoculated in scions containing a homologous transgene (Sonoda and Nishiguchi, 2000). We presume that production of tertiary siRNAs in the scion occurs because the systemic transitive signal can mark the target transcript to be used as a template by an RdRP-like activity, similar to the mechanism previously described for maintenance of RNA silencing triggered by viruses (Vaistij et al., 2002). In this context, a high expression level of the target gene, which is required to establish silencing in response to a mobile signal (Palauqui and Vaucheret, 1998), would be important to increase the probability of the signal and the target to find each other. An alternative explanation for the origin of the observed siRNAs in ZZ scions grafted onto X_Y1Y1 or XXY1Y1 rootstocks (Figure 4b, lanes 4–5) might be the transport of secondary siRNAs from rootstock to scion and their stabilization in the presence of the ZZ target. Making a sandwich graft with wild-type SR1 in-between the X_Y1Y1 rootstocks and the ZZ scions should allow to discriminate between these two possibilities. Lastly, that siRNAs in the scion would result from an interaction between the mobile silencing signal and the DNA of the target gene to produce aberrant RNAs, for instance from methylated DNA, is also unlikely because graft-induced systemic silencing seemingly occurs without induction of DNA methylation of the targeted gene (Sonoda and Nishiguchi, 2000; Mallory et al., 2003; A. Bleys and H. Van Houdt, unpublished results).

In conclusion, the most probable candidates for the mobile silencing signal are RNA amplification products. Furthermore, the signal does not seem to induce production of detectable amounts of siRNAs in the absence of the target, indicating that the systemic silencing signal is probably sent in scanty amounts and that amplification with the target RNA as a template is required to establish silencing.

Experimental procedures

Plant material

Tobacco (Nicotiana tabacum L. Petit Havana mutant SR1; Maliga et al., 1975) plants carrying a silencing-inducing locus X, a primary target locus Y (Y1 or Y2), or a combination of X and Y, were used as source of rootstocks. Wild-type SR1 tobacco plants or transgenic tobacco plants carrying a secondary target locus Z were used as source of scions. Locus X contained a T-DNA inverted repeat around the right border, in which the residing nptII-coding sequences, fused to the polyadenylation sequence (3′chs, 287 bp) of the Antirrhinum majus chs gene, were transcribed from the 35S promoter (P35S) toward the right border (Figure 1 and locus 1; Van Houdt et al., 2000). Loci Y1 and Y2 contained respectively a P35S-driven single copy of a gus-coding sequence (Figure 1 and locus Y; Van Houdt et al., 2003) and a single copy of an nptII-coding sequence (locus 2; Van Houdt et al., 2000), fused to the 3′chs terminator. Locus Z had two or more copies of a P35S-driven gus transgene carrying the 3′ end of the nos gene (De Loose et al., 1995; Van Houdt et al., 2003). Plants homozygous for locus X (XX) were crossed with plants homozygous for locus Y1 (Y1Y1) or Y2 (Y2Y2) to obtain the hemizygous hybrids X_Y1_ and X_Y2_ (the underscore indicating hemizygosity), exhibiting in trans silencing of the respective Y locus. Some of the hemizygous plants X_Y1_ were self-pollinated, and their progeny was selected on MS1 medium (Murashige and Skoog, 1962), supplemented with kanamycin or phosphinothricin to obtain XXY1_ and X_Y1Y1 plants, which were identified by segregation analysis. Progeny plants from XXY1_ and X_Y1Y1 self-pollinations were selected to contain both the X and Y1 locus and were transferred to non-selectable medium and decapitated when approximately 3 cm high: the upper part of each plant was transferred to non-selectable medium deprived of vitamins to induce root formation, and 10 days later, to soil, where the plants were grown for further use as rootstocks. Eight days after decapitation, the lower part of each plant was transferred to soil and seeds were harvested, which were subsequently sown to determine, by segregation analysis, the zygosity of the transgene loci present in the rootstock plants.

Factors involved in systemic and transitive silencing were studied in the grafts consisting of XXY1Y1, X_Y1Y1, XXY1_, or X_Y1_ gus-silenced rootstocks and ZZ gus-expressing scions. As controls, ZZ scions grafted onto XX, Y1Y1, or X_Y2_ rootstocks were used.

Grafting technique

Plants of approximately 30 and 10 cm height were used as source of rootstocks and scions, respectively (Figure 2). Both were diagonally decapitated at 5 cm from the top. One leaf below the cut in the rootstock was eliminated to facilitate attachment. For the same reason and to reduce dehydration of the scion, all the leaves in the scion longer than 10 cm were removed. The scion was fastened to the rootstock with Parafilm. The graft was covered with plastic wrap for 1 week to reduce dehydration. The lateral shoots that emerged in the rootstock after grafting were cut off to favor growth of the scion.

Enzymatic assays

β-Glucuronidase activity was determined in protein extracts, according to procedures described previously by Bradford (1976) and Breyne et al. (1993). To have a standard sampling, the extracts were made from intervenial leaf material of the third leaf below the graft junction (rootstock) and the fourth leaf above it (scion). The leaf material from the rootstocks and the scions was collected 3 and 6 weeks after grafting, respectively.

RNA gel blot analysis

Total RNA was isolated from plant samples collected 7 weeks after grafting. Each sample consisted of a mixture of the first and the second leaf below the graft junction or from the fifth to the ninth leaf above it. Total RNA was extracted with TRIzol reagent, according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA), but with double chloroform/isoamylalcohol purification. Low-molecular-weight RNA was isolated, and 25 μg of the enriched low-molecular-weight RNA fraction per lane (spectrophotometrically determined) was separated as previously described by Van Houdt et al. (2003). Predominantly ethidium bromide-stained species of low-molecular-weight RNA, separated by agarose gel electrophoresis (0.9 μg per lane) were used as loading controls. smRNA was detected as described by Van Houdt et al. (2003). The relative amount of smRNAs was calculated with the results of pixel volume quantitation (imagequant software; Amersham Biosciences, Little Chalfont, UK) of the smRNA spots.

Acknowledgements

We thank Sylvie De Buck and Gert Van der Auwera for critical reading of the manuscript and helpful comments; Annick Bleys for protein analysis; Els Van Lerberge and Anni Jacobs for technical assistance, Karel Spruyt for help with figures, and Martine De Cock for editing the manuscript. R.D.G.P. is the recipient of a doctorate fellowship from the Mexican National Council for Science and Technology (CONACYT) and the National Institute for Forestry, Agriculture and Livestock Research (INIFAP). H.V.H. is indebted to the ‘Instituut voor de Aanmoediging van Innovatie door Wetenschap en Technologie in Vlaanderen’ for a postdoctoral fellowship.

Ancillary