Viroid-induced RNA silencing of GFP-viroid fusion transgenes does not induce extensive spreading of methylation or transitive silencing

Authors


For correspondence (fax +63 21 671375; e-mail michael.wassenegger@dlr.rlp.de).

Summary

Viroid infection is associated with the production of short interfering RNAs (siRNAs), a hallmark of post-transcriptional gene silencing (PTGS). However, viroid RNAs autonomously replicating in the nucleus have not been shown to trigger the degradation of homologous RNA in the cytoplasm. To investigate the potential of viroids for the induction of gene silencing, non-infectious fragments of potato spindle tuber viroid (PSTVd) cDNA were transcriptionally fused to the 3′ end of the green fluorescent protein (GFP)-coding region. Introduction of such constructs into tobacco plants resulted in stable transgene expression. Upon PSTVd infection, transgene expression was suppressed and partial de novo methylation of the transgene was observed. PSTVd-specific siRNA was detected but none was found corresponding to the gfp gene. Methylation was restricted almost entirely to the PSTVd-specific part of the transgene. Neither a gfp transgene construct lacking viroid-specific elements was silenced nor was de novo methylation detected, when it was introduced into the genetic background of the PSTVd-infected plant lines containing silenced GFP:PSTVd transgenes. The absence of gfp-specific siRNAs and of significant methylation within the gfp-coding region demonstrated that neither silencing nor DNA methylation spread from the initiator region into adjacent 5′ regions.

Introduction

In plants, post-transcriptional gene silencing (PTGS) can be induced by transgenes (Napoli et al., 1990; Van der Krol et al., 1990), replicating viruses (Lindbo et al., 1993; Ruiz et al., 1998), and replicating satellite RNA (Gosseléet al., 2002; Wang et al., 2001). Whenever analyzed, PTGS and RNA interference (RNAi, the pendant of PTGS in animal systems) have been shown to involve the production of 21–26-bp double-stranded RNAs (dsRNAs; Hamilton and Baulcombe, 1999; Hamilton et al., 2002; Zamore et al., 2000). In both systems, these short interfering RNAs (siRNAs) derive from DICER (RNaseIII-like protein)-mediated fragmentation of dsRNA precursors (Bernstein et al., 2001). With a set of as yet unidentified proteins, siRNAs form an RNA-induced silencing complex (RISC; Hammond et al., 2000). After dissociation of one of the siRNA strands, RISC finally targets complementary RNA molecules for degradation (Nykänen et al., 2001). Based on these findings, siRNAs are recognized as a hallmark of PTGS/RNAi.

As well as being able to target RNA molecules for degradation, siRNAs can also serve as primers for cellular RNA-directed RNA polymerases (RdRPs; Sijen et al., 2001; Tang et al., 2003). During this process, the RdRP transcribes complementary RNA using the siRNA/target RNA hybrid as a template to generate secondary dsRNA. The secondary dsRNA is diced, resulting in secondary siRNAs that are homologous to sequences upstream of the initially targeted RNA region. In other words, an RISC containing secondary siRNA binds to the target RNA at a region that is located upstream of the primary siRNA-binding site. As a consequence, silencing can spread from the initiator region into 5′-adjacent sequences, a phenomenon known as transitive silencing (Vaistij et al., 2002; Van Houdt et al., 2003).

In contrast to most transgene sequences, the endogenous genes encoding phytoene desaturase (PDS) and the small subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) appeared to be protected from spreading when silencing was induced by a viral system (Vaistij et al., 2002). However, transitive silencing was also limited or absent in the case of two β-glucuronidase (GUS) transgene constructs (Elmayan et al., 1998; English et al., 1996; Wang et al., 2001). It was suggested that certain structural features of the genes or the association of transcripts with specific proteins might prevent RdRP-mediated transcription or inhibit the RdRP (Vaistij et al., 2002). On the other hand, not all endogenes are protected from transitive silencing. The endogenous tobacco β-1,3-glucanase gene was found to serve as a template for the production of secondary siRNAs (Sanders et al., 2002), although in this case, silencing was mediated by a transgene. It was proposed that endogene silencing might involve a nuclear phase that is absent during virus-induced gene silencing (VIGS).

In most cases, PTGS is associated with de novo DNA methylation, which appears to be RNA-directed (Wassenegger, 2000). We have previously shown that potato spindle tuber viroid (PSTVd) RNA directs de novo methylation of homologous genome-integrated DNA sequences (Wassenegger et al., 1994a). However, our experimental system did not allow us to determine whether the methylation was associated with PTGS (Pélissier and Wassenegger, 2000; Pélissier et al., 1999). To examine the potential of viroids for gene silencing, partial fragments of PSTVd cDNA were transcriptionally fused to the 3′ end of the green fluorescent protein (GFP) gene. These constructs were introduced into tobacco, and GFP activity was monitored in non-infected and PSTVd-infected plants. As expected (Itaya et al., 2001; Papaefthimiou et al., 2001), viroid-specific siRNAs were produced only in infected plants. In addition to siRNA production, we could show that infection resulted in the inactivation and partial de novo methylation of the chimeric GFP:PSTVd transgenes. The gfp gene is prone to transitive silencing over its entire length irrespective of whether silencing is induced by a transgene or a virus (Vaistij et al., 2002; Voinnet et al., 1998, 2000). Therefore, we expected that PSTVd-induced silencing would also initiate transitive and trans-silencing of the gfp gene in this system. To test this hypothesis, a second gfp transgene construct, lacking any viroid-specific sequence, was introduced into transgenic plants containing silenced GFP:PSTVd transgenes. If silencing could spread from the PSTVd-specific part of the construct into the upstream gfp sequence, expression of the second gfp transgene should be co-suppressed. However, it turned out that silencing was restricted to transgenes containing the PSTVd sequence. The gfp transgene, lacking PSTVd sequences, was expressed normally and remained unmethylated. Moreover, no gfp-specific siRNAs were detected demonstrating that silencing did not spread from the PSTVd region into upstream gfp sequences.

Results

Introduction of chimeric GFP:PSTVd transgenes into tobacco plants

Two GFP:PSTVd transgenes were constructed by introducing fragments of PSTVd cDNA into the binary plant transformation vector pPCV702SM. The cDNA fragments were 98 bp (PSTVd98) and 160 bp (PSTVd160) in length, giving rise to constructs pPCV702PSTVd98 and pPCV702PSTVd160, respectively. The entire gfp cDNA was then inserted yielding recombinant constructs 702GFP98 and 702GFP160 (Figure 1). In addition to the hybrid GFP:PSTVd constructs, the gfp-coding region was inserted into the expression cassette of vector pPCV702SM to generate a viroid-free gfp transgene construct (702GFP, Figure 1). All constructs were introduced into tobacco SR1 plants via Agrobacterium-mediated leaf disk transformation. Ten independent transformants representing each construct were regenerated, resulting in three construct groups (SR1GFP98, SR1GFP160, and SR1GFP). These were screened for the presence of the transgene by Southern blot hybridization (data not shown).

Figure 1.

Physical map of the 702GFP, 702GFP98, and 702GFP160 transgene constructs.

Restriction sites that were relevant for Southern blot analysis are indicated. The cauliflower mosaic virus P35S, the GFP, and the polyadenylation signal (pA) sequences are not shown to scale as indicated by the vertical broken lines. Double diagonal lines indicate that the most upstream and downstream DraI and Bsp119I fragments are truncated for clarity. LB, T-DNA left border sequence; RB, T-DNA right border sequence.

UV illumination and Northern blot hybridization were used to monitor GFP fluorescence and to check steady-state gfp mRNA levels. These procedures revealed that all transgenic lines expressed the gfp transgenes (data not shown). In tobacco plants, transcription of the 98- and 160-bp PSTVd cDNA fragments does not result in viroid infection (Wassenegger et al., 1994b, 1996). This was confirmed by Northern blot analysis of total RNA isolated from the SR1GFP98 and SR1GFP160 plants using a PSTVd-specific probe. No hybridization signal corresponding to replicating PSTVd RNA was detected, showing that none of the transgenic plant lines was infected with the viroid (data not shown). Two transgenic lines were selected from each of the three construct groups (SR1GFP98, SR1GFP160, and SR1GFP), one containing a single copy transgene and one containing multiple copies (up to a maximum of four). These six plants were self-pollinated to produce homozygous lines.

PSTVd-infection of SR1GFP98 and SR1GFP160 plant lines

Potato spindle tuber viroid infection was achieved by crossing the homozygous SR1GFP98, SR1GFP160, and SR1GFP plants with the homozygous SR1-4(−) plant line containing the 702PSTVd-4(−) construct (see Figure 6) comprising four tandemly linked full-length copies of the PSTVd cDNA. It has been shown previously that all tobacco plants expressing multimeric full-length PSTVd cDNAs become infected with the viroid (Wassenegger et al., 1994a). Nevertheless, Northern blot analysis was carried out using total RNA isolated from progeny plants and a PSTVd-specific probe to confirm our previous data. High concentrations of the mature PSTVd RNA were detected showing autonomous replication of the PSTVd RNA in all plants (data not shown). In addition to these genetic crosses, 10 SR1GFP98 plants were infected with PSTVd by mechanical inoculation using total RNA from the SR1-4(−) line. Northern blot analysis showed that two of the inoculated plants were infected (data not shown).

Viroid-induced RNA silencing (VdIRS)

In a first set of experiments, GFP activity in the PSTVd-infected plants was examined under UV-light. We found that GFP fluorescence was strongly reduced in SR1GFP98 × SR1-4(−) (Figure 2a1) and SR1GFP160 × SR1-4(−) (data not shown) progeny plants. In contrast, GFP activity was not reduced in SR1GFP × SR1-4(−) plants (Figure 2b), indicating that silencing was targeted to the PSTVd region of the transgene mRNA. The fact that mechanically inoculated SR1GFP98 lines also lost GFP fluorescence (Figure 2a2) showed that the silencing effect was not based on the presence of the PSTVd-4(−) transgene but was because of PSTVd infection.

Figure 2.

Viroid-induced silencing of the GFP-PSTVd transgene.

Suppression of gfp expression in PSTVd-infected SR1GFP98 plants was monitored under UV-light where GFP-expressing plants show green and non-expressing plants show red fluorescence. The SR1GFP98 × SR1-4(−) plant line (a1) was infected through the introduction of an infectious transgene construct whereas the SR1GFP98inoc. plant (a2) was mechanically inoculated. SR1GFP plants carrying a gfp transgene unlinked to the PSTVd cDNA fragment were not silenced upon PSTVd infection (b).

To provide further evidence that viroid infection induced GFP: PSTVd transgene silencing, the steady-state level of gfp mRNA was measured. Northern blot analysis using the gfp cDNA as a probe revealed that transgene mRNA was almost absent in all SR1GFP98 × SR1-4(−) (Figure 3a, lanes 2 and 4) and SR1GFP160 × SR1-4(−) lines (data not shown). No significant difference in gfp mRNA levels was found in infected SR1GFP × SR1-4(−) lines when compared with viroid-free SR1GFP lines (Figure 3a, lanes 5 and 6). These data indicated that VdIRS was extremely efficient. Additional evidence for VdIRS was obtained by analyzing the population of small RNAs in the infected and non-infected plants. As was previously shown by others (Itaya et al., 2001; Papaefthimiou et al., 2001), we also found that PSTVd infection was associated with the production of viroid-specific siRNAs. Irrespective of whether infection was induced by a transgene or by mechanical inoculation, approximately 23-bp-long siRNAs were present in all PSTVd-infected plants (Figure 4a). By contrast, Northern blot hybridization under similar conditions, but with the GFP probe that included the region neighboring the PSTVd insertion, revealed that no GFP-specific siRNAs were produced (Figure 4b).

Figure 3.

Northern blot analysis of steady-state gfp mRNA levels.

(a) Total RNA isolated from SR1GFP98/1b and SR1GFP98/7b (lanes 1 and 3), SR1-4(−)/6 × SR1GFP98/1b and SR1GFP98/7b (lanes 2 and 4), SR1GFP/14b (lane 5), and SR1GFP/14b × SR1-4(−)/6 (lane 6) plant lines was hybridized against the entire gfp cDNA as probe.

Hybridization signals in lanes 2 and 4 were hardly detectable indicating that the GFP:PSTVd98 transgene was silenced in infected SR1GFP98 plant lines carrying either a single-copy transgene (lane 2) or multiple transgene copies (lane 4).

(b–d) Re-hybridization of the filter with a PSTVd cDNA probe (b) demonstrated that all plants containing the PSTVd-4(−) transgene were PSTVd-infected. Photographs of the agarose gel (c) and the corresponding blot (d) showed that almost equal amounts of RNA were electrophoretically separated and transferred onto the membrane.

Figure 4.

Northern blot analysis to detect siRNAs.

(a) Detection of PSTVd-specific siRNAs. Northern blot analysis of low-molecular weight RNA isolated from SR1 wild type (lane 1), SR1-4(−)/6 (lane 2), SR1GFP98/7b (infected by mechanical inoculation; lanes 3 and 4) and SR1-4(−)/6 × SR1GFP98/1b and SR1GFP98/7b (lanes 5 and 6) plant lines was carried out using a PSTVd cDNA probe. The lower hybridizing fragments corresponded to sizes of 22–25 nt.

(b) Absence of GFP-siRNA accumulation. An 1.5% agarose gel is shown (I) to demonstrate that almost equal amounts of low-molecular weight RNAs were electrophoresed and that all RNAs were of high quality. For a Northern blot experiment, RNAs were separated on a 15% polyacrylamide (PAA) gel, blotted onto a nylon membrane and hybridized with a GFP cDNA probe (II). Total RNAs were isolated from SR1-4(−)/6 (lane 1), SR1GFP/15E (lane 2) (SR1GFP98/5b × SR1-4(−)/6)-A1 (lane 3) plants, two progeny of the triple cross (SR1GFP98/5b × SR1-4(−)/6)-A1 × SR1GFP/15E (lanes 4 and 5) (SR1GFP98/14b × SR1-4(−)/6)-A1 plant (lane 7) and four progeny of the triple cross (SR1GFP98/14b × SR1-4(−)/6)-A1 × SR1GFP/15E (lanes 8–11). 5′ end-labeled 21- and 23-nt-long oligonucleotides were used as molecular weight markers (lane 6).

De novo DNA methylation

From our previous experiments, we were aware that genome-integrated PSTVd cDNA sequences ≥30 bp in size are efficiently de novo methylated upon PSTVd infection (Pélissier and Wassenegger, 2000). Therefore, we expected and confirmed that the viroid-specific parts of the GFP:PSTVd transgenes were methylated in the infected SR1GFP98 × SR1-4(−) and SR1GFP160 × SR1-4(−) plants. Southern blot analysis revealed that the unique HpaII site present within the viroid cDNA in both the 702GFP98 and 702GFP160 constructs was heavily methylated in the infected plants (Figure 5, lanes 2 and 4). To avoid any cross-hybridization with the SR1-4(−) transgene, the blot was hybridized with a gfp probe. The appearance of only one additional fragment (Figure 5, 1 kbp in lane 2 and 1.06 kbp in lane 4) showed that HpaII sites flanking the unique site within the PSTVd sequence were non-methylated in both plant lines. Thus, methylation appeared to be restricted to the viroid-specific part of the transgenes.

Figure 5.

Southern blot analysis with DraI/HpaII-digested genomic DNA was carried out to examine de novo methylation patterns of the PSTVd-specific part of the GFP:PSTVd98 (lane 1) and GFP:PSTVd160 (lane 3) transgenes.

When probed with the entire gfp cDNA, only one 860-bp fragment was expected to hybridize if the DNA was not methylated and completely cleaved by HpaII (see physical map above the autoradiograph). The viroid-free SR1GFP98/5b (lane 1) and SR1GFP160/8b (lane 3) plants displayed a corresponding pattern, illustrating that the DNA was not modified. The faint band indicated by an arrow derived from cross-hybridization with the T-DNA. This fragment also lit up when genomic DNA from a transgenic plant line carrying the T-DNA of the empty pPCV702 vector was used (lane 5). The major part of the approximately 860-bp HpaII fragment shifted to 1000 bp (lane 2) and 1060 bp (lane 4) when SR1GFP98/5b × SR1-4(−)/6 and SR1GFP160/8b × SR1-4(−)/6 DNA was analyzed, respectively. These sizes were expected for methylated copies of the transgenes.

The downstream HpaII site lies about 60 bp from the HpaII site in the PSTVd sequence and is found within the transgene's polyadenylation signal. This site was always cleaved, showing that if 3′ spreading of methylation was taking place, it did not extend more than 60 bp from its site of origin. There are no HpaII sites in the gfp cDNA, so analysis of de novo methylation in the near upstream sequences was not possible using HpaII. We therefore digested genomic DNA from SR1GFP98 × SR1-4(−) progeny plants with Bsp119I and EcoRI, and carried out Southern blots using a 1780-bp Bsp119I fragment of plasmid 702GFP98 as the probe. A unique Bsp119I site (5′-TTCGAA-3′) is located within the gfp sequence and a second is found within the neomycin phosphotransferase II (nptII) gene of the binary vector. The gfp-specific Bsp119I site is located only 100 bp upstream of the gfp–PSTVd junction (Figures 1 and 6) and, like HpaII, Bsp119I is sensitive to cytosine methylation. The expected sizes of hybridizing fragments corresponding to fully digested non-methylated DNA were 1780 bp (GFP:PSTVd98) and 3150 bp (PSTVd-4(−)). This pattern was found for most of the SR1GFP98 × SR1-4(−) progeny plants (Figure 6, lane 2). Nevertheless, the appearance of a 2490-bp fragment showed that the gfp-specific Bsp119I site was partially methylated in some of these plant lines (Figure 6, lanes 1 and 3). In one line, remarkably high levels of methylation were found at this Bsp119I site (Figure 6, lane 1), suggesting occasional 5′ spreading of methylation. However, significant methylation was exceptional and was never observed in plants containing the gfp, GFP:PSTVd98, and PSTVd-4(−) transgene constructs (see below).

Figure 6.

Southern blot analysis with EcoRI/Bsp119I-digested genomic DNA was carried out to examine de novo methylation patterns of the GFP:PSTVd98 transgene.

The DNA of three progeny of the self-pollinated SR1GFP98/5b × SR1-4(−)/6 plant line was hybridized using the 702-Bsp119I fragment as the probe. A single 1780-bp band was expected if the GFP:PSTVd98 transgene was not methylated and completely cut by Bsp119I (see physical map above the autoradiograph). The additional approximately 3150-bp band originated from the PSTVd-4(−) transgene (see most upper physical map). The appearance of an approximately 2490-bp fragment (lanes 1 and 3) indicated that the Bsp119I site within the gfp sequence was partially methylated in the (SR1GFP98/5b × SR1-4(−)/6)-A1 and (SR1GFP98/5b × SR1-4(−)/6)-A3 plant lines.

Transitive silencing

Homozygous SR1GFP plants were crossed with the SR1GFP98 × SR1-4(−) and SR1GFP160 × SR1-4(−) lines to produce plants containing all three transgene constructs. Progeny were pre-screened for the presence of the transgenes by PCR. Positive plants were further analyzed by Southern blot hybridization using EcoRI/Bsp119I-digested genomic DNA. The blot was hybridized using the 702-Bsp119I fragment as a probe, allowing the detection of all three transgene constructs in parallel. The hybridizing fragment was 1680 bp in size and showed that the gfp transgene was present (Figure 7, lanes 1, 3, and 4). The 1780-bp fragment (Figure 7, lanes 1 and 2) and the 1840-bp fragment (Figure 7, lanes 3–5) were diagnostic for the GFP98 and GFP160 transgenes, respectively. The 3150-bp fragment (Figure 7, lanes 1 and 3) corresponded to the SR1-4(−) transgene. Importantly and in contrast to SR1GFP98 × SR1-4(−) progeny plants (Figure 6), no methylation of the Bsp119I site was detectable in progeny plants of the triple crosses.

Figure 7.

Southern blot analysis of triple crossed plant lines to verify the presence of all three transgene copies.

EcoRI/Bsp119I-digested genomic DNA was probed using the 702-Bsp119I fragment that recognized all three transgene constructs. The approximately 1680-bp band corresponded to the Bsp119I fragment of the gfp transgene, the approximately 1780-bp band corresponded to the Bsp119I fragment of the GFP:PSTVd98 transgene and the approximately 1840-bp band corresponded to the Bsp119I fragment of the GFP:PSTVd160 transgene. Lane 1: (SR1GFP98/5b × SR1-4(−)/6)-A1 × SR1GFP/14b line 1; lane 2: SR1GFP98/5b; lane 3: (SR1GFP160/8b × SR1-4(−)/6)-A1 × SR1GFP/15E line 1; lane 4: SR1GFP160/8b × SR1GFP/14b; and lane 5: SR1GFP160/8b.

Northern blot analysis of the plants revealed that the gfp gene was strongly expressed (Figure 8). The presence of a silenced GFP98 (Figure 8, lanes 3 and 4) or GFP160 transgene (Figure 8, lanes 5 and 6) had no obvious effect on the expression of the viroid-free gfp transgene. Conversely, introduction of the active, viroid-free gfp transgene appeared to have no impact on the silenced PSTVd-containing gfp transgenes. Hybridization against the GFP probe demonstrated that the GFP98 and GFP160 transcripts were absent in the triple crossed plants where the GFP transgene was expressed (Figure 8a). Based on the size differences between the GFP and GFP160 transcripts, at least these two mRNAs could be clearly discriminated on a Northern blot (Figure 8e). It should be noted that PSTVd probes strongly hybridized with PSTVd replication intermediates and were therefore not suitable to specifically detect transcripts of the GFP:PSTVd98 and GFP:PSTVd160 transgenes in infected plant lines (Figure 9). There were no significant differences in gfp hybridization signal strength, when SR1GFP × SR1-4(−) (Figure 8a, lane 2) was compared with two progeny plants of the SR1GFP98 × SR1-4(−) × SR1GFP (lanes 3 and 4) and SR1GFP160 × SR1-4(−) × SR1GFP triple crosses (lanes 5 and 6), respectively. Similar to the data we obtained with the double crosses, Northern blot analysis of small RNAs isolated from triple crosses using the GFP probe failed to detect GFP-specific siRNAs (Figure 4b).

Figure 8.

Northern blot analysis of steady-state gfp mRNA levels.

(a) Total RNA isolated from individual plants of lines SR1-4(−)/6 (lane 1), SR1GFP/14b × SR1-4(−)/6 (lane 2), two progeny of the triple cross (SR1GFP98/5b × SR1-4(−)/6)-A1 × SR1GFP/14b (lanes 3 and 4) and two progeny of the triple cross (SR1GFP160/8b × SR1-4(−)/6)-A1 × SR1GFP/14b (lanes 5 and 6) was hybridized with a probe comprising the entire gfp cDNA.

No significant differences between the SR1GFP × SR1-4(−) and the triple crossed plants were detectable indicating that the gfp transgene was moderately expressed despite the presence of a silenced GFP:PSTVd transgene. The plain and dotted arrows indicate the expected positions of the GFP160- and GFP98-specific transcripts (see below).

(b–d) Re-hybridization of the filter with the PSTVd cDNA probe (b) demonstrated that all plants were PSTVd-infected. Photographs of the agarose gel (c) and the corresponding blot (d) showed that almost equal amounts of RNA were electrophoretically separated and transferred onto the membrane.

(e) Northern blot obtained with RNA of the SR1GFP160/8b (lane 1), SR1GFP160/8b × SR1GFP/15E (lane 2) and SR1GFP/15E (lane 3) plant lines using the entire gfp cDNA as probe. The corresponding hybridization patterns showed that GFP160 and GFP transgene transcripts can be discriminated by Northern blot analysis.

Figure 9.

Northern blot analysis to detect GFP-PSTVd transgene expression using a PSTVd-specific probe.

Total RNA isolated from the SR1GFP98/5b (lane 1) and SR1GFP98/5b × SR1-4(−)/6 (lane 2) plant lines was hybridized with PSTVd cDNA as probe. Only a faint band hardly detectable on the autoradiograph could be visualized in lane 1, representing the GFP98 transgene transcript. In PSTVd-infected plants, the presence of numerous PSTVd replication intermediates precluded the detection of the GFP98- and GFP160-specific transcripts. The positions of the mature PSTVd RNA form and of the GFP98 mRNA are indicated on the left side of the figure.

Discussion

Induction of PTGS upon PSTVd infection

We have shown here that mRNAs with 3′ homology to PSTVd sequences are specifically and efficiently targeted for silencing upon PSTVd infection. Silencing correlated with the accumulation of approximately 23-bp-long PSTVd-specific siRNAs and with methylation of the 3′ transcribed part of the GFP:PSTVd transgenes, two hallmarks of PTGS. Methylation was restricted to the viroid counterpart of the transgenes. Importantly, no methylation was found in the cauliflower mosaic virus 35S promoter (P35S) sequence of the GFP98 and GFP160 transgene constructs, which strongly argued against a contribution of TGS (Paszkowski and Whitham, 2001; Sijen et al., 2000). Moreover, out-crossing of the PSTVd-4(−) locus resulted in loss of PSTVd infection and in complete reactivation of GFP98 expression (data not shown). This contrasts with our current understanding of TGS. In progeny plants that lost the silencer locus upon out-crossing, transcriptional silencing of transgenes was found to be maintained (Paszkowski and Whitham, 2001). In addition, hypermethylation of coding regions was associated with a wide range of genes that were undergoing PTGS, including the gfp gene and, to date, no evidence could be provided that indicated an impact of this coding region methylation on transcription in higher eukaryotes. Therefore, presence of PSTVd-specific siRNAs, specific de novo methylation of only the PSTVd part of the transgenes and the apparent re-expression of the GFP98 transgene after meiosis in viroid-free progeny all pointed to a PTGS process. For this process, siRNA production would be initiated upon PSTVd infection. The PSTVd-specific siRNAs accumulating in the cytoplasm would then target the GFP98 or GFP160 transgene transcripts for degradation at the PSTVd-specific part. Similar to PTGS that was observed for the endogenous PDS and RuBisCO genes (Vaistij et al., 2002) and a GUS transgene construct (English et al., 1996) spreading of silencing into upstream regions was not detectable. These findings may indicate that induction or inhibition of transitive silencing requires yet to be identified features that could be determined by the sequences context.

Origin of PSTVd-specific siRNAs

In contrast to RNA viruses, PSTVd replicates in the nucleus. PSTVd replication proceeds via the RNA/RNA pathway, a mechanism that involves the synthesis of multimeric (+)- and (−)-specific RNA genomes (Sänger, 1987). The (+) and (−) RNA strands have the potential to form dsRNA but it is not yet clear whether this actually happens in the nucleus. If nuclear dsRNA can be formed, one has to presume that the siRNAs produced in PSTVd-infected plants are the result of nuclear DICER activity. This is in agreement with Mette et al. (2000, 2001) who expressed in tobacco plants an inverted repeat (IR) transgene construct lacking a polyadenylation signal. The non-polyadenylated double-stranded primary transcript of the IR transgene was retained in the nucleus. Nevertheless, the RNA was partially cleaved into siRNAs demonstrating that a nuclear DICER exists in plants (Mette et al., 2001; Papp et al., 2003). Importantly, and in contrast to our data, Mette et al. (2001) reported that the nuclear-derived siRNAs could not trigger substantial degradation of a homologous mRNA. Thus, we do not exclude that PSTVd-specific siRNAs are exclusively produced in the cytoplasm or are produced in both the cytoplasm and the nucleus. It is conceivable that dsRNA replication intermediates of the viroid are also present in the cytoplasm. Efficient processing by DICER may explain the failure to detect them. Notably, recent findings demonstrated that PSTVd-specific siRNAs accumulated in the cytoplasm and were hardly detectable in the nucleus (M. Tabler and coworkers, unpublished results). This observation indicates that the siRNAs are either produced in the cytoplasm or processed in the nucleus from where they are rapidly and efficiently exported into the cytoplasmic compartment.

Transgene-specific de novo methylation

In viroid-infected SR1GFP98 and SR1GFP160 plants, the pattern of de novo methylation detected within the integrated PSTVd cDNA sequences appeared to be similar to that detected previously in viroid-infected plants containing only the pPCV702-PSTVd98 or pPCV702-PSTVd160 constructs (Pélissier and Wassenegger, 2000). In these transgene constructs, the PSTVd cDNAs were fused directly to the P35S. By contrast, in the present study, the PSTVd cDNAs were linked to an upstream gfp-coding region that was transcribed in the SR1GFP98 and SR1GFP160 plant lines. The fact that in both experiments almost identical methylation patterns were established showed that transcription of the upstream sequences had no apparent effect on the extent or efficiency of de novo methylation (see above; Pélissier and Wassenegger, 2000). Considering earlier reports showing that in silenced plants, methylation could spread along a transcribed gfp sequence (Jones et al., 1999), this finding indicated that transitive silencing was not induced in our system (see below).

Southern blot analysis using genomic DNA cleaved with Bsp119I revealed that, in a few infected plant lines, the 3′ end of the gfp-coding region showed low levels of de novo methylation. In one plant ((SR1GFP98/5b × SR1-4(−)/6)-A1), non-cleaved DNA was found in about 10% of the leaf cells (Figure 6). Whether this methylation occurred because of migration or spreading could not be determined. Migration is thought to reflect error prone methylase activity, and involves the DNA methylase sliding beyond its target region, in this case PSTVd cDNA, into adjacent DNA sequences. This process does not require a secondary inducer. In contrast, spreading is dependent on an inducer, specifically an RNA molecule that would direct the methylase to upstream gfp sequences. Unfortunately, the Bsp119I site at the 3′ end of the gfp gene was never methylated in any of our triple crosses. It should be noted that gfp gene methylation was also absent in progeny plants of a genetic cross between the (SR1GFP98/5b × SR1-4(−)/6)-A1 and the SR1GFP lines (Figure 7, lane 1). If this site had been methylated, it would have been interesting to determine whether only the GFP:PSTV transgenes were modified, or whether the gfp transgene was also affected. The former would have supported the migration model, whereas the latter would indicate the presence of a specific inducer capable of functioning in trans, and would have supported the spreading model.

Transitive silencing

Potato spindle tuber viroid infection efficiently silenced the GFP98 and GFP160 transgenes, but there was no trans-silencing as demonstrated by the expression of the viroid-free gfp transgene in infected SR1GFP98 × SR1GFP and SR1GFP160 × SR1GFP plants (Figure 8). The absence of secondary gfp-specific siRNAs was in accordance with these observations. Secondary siRNAs are supposed to derive from secondary dsRNA whose production is assumed to result from transcription of heavily methylated sequences (Voinnet et al., 1998) or from RdRP-mediated transcription of siRNA-primed target RNA (Sijen et al., 2001). Heavy methylation within coding region of post-transcriptionally silenced genes was suggested to lead to a certain level of premature termination of transcription or may impose a yet to be defined tag to the resulting transcripts, allowing de novo formation of aberrant RNA molecules. In wheat extracts, RdRP-mediated dsRNA synthesis from non-primed templates was demonstrated (Tang et al., 2003), supporting the idea that dsRNA could be produced using aberrant primary gene transcripts that derived from transcription of methylated DNA. RdRP-mediated production of secondary dsRNA from siRNA-primed target RNA is evident in animal and plant systems (Sijen et al., 2001; Tang et al., 2003). However, although the GFP98 and the GFP160 transgenes were heavily methylated at their 3′ ends and PSTVd-specific siRNAs were produced upon viroid infection, theoretically having the potential to prime secondary dsRNA transcription, we could not detect any secondary siRNAs (Figure 4b). This finding suggested that neither unprimed nor primed synthesis of secondary dsRNAs took place in our system.

Viroids are self-complementary RNA molecules and can form a variety of secondary structures (Gruner et al., 1995). We speculate that the PSTVd98 and PSTVd160 RNAs may form secondary structures that block RdRP-mediated transcription. It should be noted that both partial PSTVd RNAs have identical 5′ end sequences, which probably form identical structures. It is possible that these structures form ‘bookends’ that the RdRP cannot remove. It may be possible to determine the precise sequences responsible for blocking transitive silencing using cDNA fragments that cover other regions of the PSTVd genome. In addition, the 5′ end of the PSTVd160 sequence can be stepwise shortened to obtain alternative RNA 5′ structures. We further plan to utilize inverted repeat constructs and chimeric viruses to target the PSTVd-specific parts of the GFP98 and GFP160 transgenes for silencing. Exploitation of these alternative silencing inducers will show whether the failure of initiating transitive silencing is specific of the VdIRS system.

The failure of VdIRS to initiate transitive silencing is reminiscent of the satellite RNA system where the spreading of silencing into upstream sequences could also not be detected (Wang et al., 2001). However, there are two major differences between the satellite and PSTVd silencing systems. First, Wang and coworkers transcriptionally fused a satellite cDNA fragment to the 3′ end of the gusA gene. At the time, it was not known that the very 3′ end (last 200 nt) of the gusA mRNA is a weak target for VIGS (Braunstein et al., 2002). Moreover, no 5′ spreading of silencing could be detected along the gusA sequence. This may explain the failure of the satellite system to induce transitive silencing. Second, as is the case for RNA viruses, satellite RNA replicates in the cytoplasm whereas PSTVd replication takes place in the nucleus. It has been shown that cytoplasmic inducers (viruses, satellites) are less potent than nuclear inducers (transgenes), at least for the silencing of endogenous genes (Sanders et al., 2002). Therefore, one may speculate that nuclear-replicating viroids would provide the component that is required for a nuclear silencing step. However, our data suggest that neither the context of the target sequence nor a putative nuclear factor was sufficient to trigger transitive silencing. Although, in numerous studies, the gfp gene was shown to be highly susceptible to PTGS and transitive silencing in both the 5′-and 3′ directions (Vaistij et al., 2002; Voinnet et al., 1998), we neither observed spreading of methylation into upstream gfp sequences nor transitive GFP silencing in the SR1GFP98 × SR1-4-(−) × SR1GFP and SR1GFP160 × SR1-4(−) × SR1GFP plants.

Experimental procedures

Plasmid construction

The PSTVd98 cDNA fragment was amplified by PCR from a plasmid containing the cDNA of a PSTVd mutant using the Bgl67-H forward (5′-TTAGATCTCCGCTTTTTCTC-3′) and the PV30-H reverse (5′-TGAACCACAGGAACCACGAGGGGGG-3′) primers according to Wassenegger (2001). The PCR product was inserted into the pTPCR vector (Wassenegger et al., 1994b). From this plasmid, the PSTVd98-specific fragment was released by HindIII digestion and inserted into the unique HindIII site of pPCV702SM (Wassenegger et al., 1994b). The PSTVd mutant carried four nucleotide substitutions (T7 → C, A8 → C, A9 → C, and A10 → C) to distinguish it from the original sequence of the PSTVd KF440-2 isolate. For construction of the final 702GFP98 vector, the gfp BamHI/SacI cDNA fragment with trimmed SacI overhang was ligated into the BamHI/XbaI sites of pPCV702PSTVd98 (XbaI partially filled).

The PSTVd160 cDNA fragment was amplified by PCR from a plasmid containing the cDNA of PSTVd KF440-2-type strain using the Bgl67-H forward and the PV175-H reverse (5′-TCCTGTCGGCCGCTGGGCACTCCC-3′) primers. The PCR product was cleaved with BglII and BamHI, and inserted into the BamHI site of vector pT3T7 (Boehringer, Mannheim, Germany) to give pT3T7PSTVd160. A HindIII/BamHI fragment containing PSTVd160 was isolated from this plasmid, the BamHI overhang filled, and the fragment was inserted into the HindIII/SalI sites of pPCV702SM (SalI ends filled), resulting in pPCV702PSTVd160.

The 702GFP160 vector was obtained by insertion of the gfp BamHI/SacI cDNA fragment (SacI end trimmed) into the BamHI/BglII sites of pPCV702PSTVd160 (BglII end filled). The recombinant 702GFP was produced by the insertion of an approximately 820-bp BamHI/SacI (trimmed) fragment of gfp cDNA into the BamHI/BglII (filled) sites of pPCV702SM.

Plant tissue culture and transformation

The SR1-4(−) plant line was generated as described previously by Wassenegger et al. (1994a). Nicotiana tabacum (cv. Petit Havana SR1) plants used for transformation were maintained and propagated under sterile conditions. Leaf disk transformation with the pPVC702SM derivatives, selection, and regeneration of transformants was performed according to the procedure described by Wassenegger et al. (1994b). Cuttings of the transformants were transferred into soil and maintained in a greenhouse under standard conditions (10–14 h daylight, 22–24°C). Homozygous transformants were obtained by self-pollination. Progeny were tested for the homozygous state of the transgene(s) by Southern blot analysis and by backcrossing to SR1 wild-type plants. Genetic crosses were performed in the greenhouse by emasculating young buds of homozygous transformants (SR1GFP, SR1GFP98, and SR1GFP160) and by dusting pollen of either wild-type (backcross) or homozygous transformants (SR1-4(−)) on the stigmata. Progeny were tested for the presence of the transgenes by PCR and Southern blot analysis. Accordingly, triple crosses were obtained using double crosses as parental plants.

PCR analysis

Polymerase chain reaction amplification of genome-integrated transgenes was carried using 1 µg of DraI-digested genomic DNA in a total volume of 100 µl. The reaction mix was made up according to the manufacture's instructions (TaKaRa) and 1.5 U of Taq DNA polymerase was used (TaKaRa Ex Taq). Gene-specific fragments were amplified under the following conditions: 1 cycle at 95°C for 2 min; 30 cycles at 95°C for 40 sec, 58°C for 2 min, 72°C for 3 min; and 1 cycle at 72°C for 10 min. For simultaneous amplification of the gfp (702GFP), PSTVd (SR1-4(−)), and GFP:PSTVd98 (702GFP160) transgenes, we used primers specific for the polyadenylation signal (1020BiUSrev: 5′-CTCTAATCATAAAAACCCATCTC-3′) and cauliflower mosaic virus P35S (710BiUS-35S: 5′-AAGCAAGTGGATTGATGTG-3′). PCR products were analyzed on a 1.5% agarose gel.

DNA isolation and Southern blot analysis

Genomic DNA was extracted from plants according to Dellaporta et al. (1983). We digested 10–15 µg of genomic DNA overnight with appropriate restriction enzymes (Roche, Mannheim, Germany; Fermentas, St. Leon–Rot, Germany). DraI/HpaII- and EcoRI/Bsp119I-digested genomic DNA was separated on 1.2 or 0.8% agarose gels at 60 V overnight. For DNA transfer onto positively charged nylon membranes (Qiagen, Hilden, Germany) agarose gels were soaked once in 0.2 m HCl for 10 min, once in 0.5 m NaOH, 1.5 m NaCl for 45 min, and twice in 0.5 m Tris (pH 7.2), 1.5 m NaCl, and 1 mm EDTA for 15 min. Capillary transfer was performed overnight according to Sambrook et al. (1989). DNA was UV cross-linked and processed for hybridization according to Amasino (1986) using random primed [α-32P] dCTP-labeled DNA (Ladderman, Labelling kit, Takara.Mirius.Bio, Madison, WI, USA) as the probe. Hybridizing DNA fragments were identified by autoradiography after exposure at −80°C for 16–72 h.

RNA isolation and Northern blot analysis

Total RNA was extracted from leaves in 8 m guanidine–HCl buffer (pH 7.0) according to the method described by Logemann et al. (1987). RNA (10–15 µg) was separated on formaldehyde-containing agarose gels (Sambrook et al., 1989) and treated prior to overnight capillary transfer onto uncharged nylon membranes (Qiagen) with 20× SSC, 0.05 m NaOH, and 1.5 m NaCl for 20 min (once), with 0.5 m Tris (pH 7.4), 1.5 m NaCl for 15 min (twice), and with diethylpyrocarbonate (DEPC) water for 5 min (once). RNA was immobilized onto the membrane and successful transfer was monitored by taking a photograph of the membrane under UV light. Hybridization was performed overnight at 42°C in a formamide-containing buffer with random primed [α-32P] dCTP-labeled DNA probes (Ladderman). Hybridizing RNA was identified by autoradiography after exposure at −80°C for 16–72 h.

GFP visualization

Green fluorescent protein activity in transgenic plants was monitored under UV light with a NICON COOLPIX990 digital camera using the Heliopan Yellow 8 ES 62 light filter. Camera settings were: artificial light, matrix, flashlight (off), ASA 400, fine, aperture = 2, exposure time = 2.9 sec, and manual focus. The pictures were taken with a hand-held UV lamp (Black–Ray Long Wave UV lamp, Model B 100 AP) with a distance to the object of 40–70 cm.

siRNA detection

Detection of siRNAs was performed according to Hamilton and Baulcombe, 1999) and Mette et al. (2000), respectively. Total RNA was extracted from young plant leaves (frozen in liquid nitrogen) with TRIzol reagent (Life Technologies, Rockville, MD, USA) according to the manufacturer's instructions. RNA samples were dissolved in deionized formamide, separated on 15% (v/w) polyacrylamide gels containing 7 m urea, and transferred onto uncharged nylon membranes (Qiagen) by vertical electrophoreses. We synthesized [α-32P] dUTP-labeled RNA probes in vitro from linearized plasmid DNA using a T3/T7 transcription kit (Promega, Mannheim, Germany). Probes corresponding to (+) and (−) transcripts of the entire GFP cDNA and of the PSTVd cDNA sequences were hydrolyzed into approximately 50-nt fragments according to Mette et al. (2000).

Mechanical viroid infection and detection

Mechanical viroid inoculation was performed as previously described by Wassenegger et al. (1996). Total RNA isolated from SR1-4(−) plants was used as the inoculum. Infection was monitored by Northern blot analysis as described above or detection of mature PSTVd RNA using glyoxalated RNA separated by polyacrylamide gel electrophoresis, blotted, and probed with [α-32P]-CTP-labeled PSTVd cDNA (Spiesmacher et al., 1985).

Acknowledgements

We thank Richard M. Twyman for helpful comments on the manuscript and suggestions for its improvement. Our research was supported by the Deutsche Forschungsgemeinschaft (grants: Wa1019/1-1-1-4) and the Fraunhofer Institute for Molecular Biology and Applied Ecology.

Ancillary