In plants, post-transcriptional RNA silencing (PTGS) signals move beyond their sites of synthesis using channels called plasmodesmata (PD). However, a potential influence of PD permeability in silencing signal movement has not been addressed previously. PD connectivity and aperture are routinely determined by monitoring the cell-to-cell movement of protein tracers of different sizes. Here we compare protein and RNA silencing signal movement during Arabidopsis embryogenesis. The data suggest that the degree of PTGS signal movement correlates with the degree of PD aperture previously determined in distinct subregions of the developing Arabidopsis embryo. Silencing signals move more extensively in the hypocotyl and root compared with the area around the tips of the cotyledons, suggesting that the site of silencing signal initiation influences the extent of signal movement. In addition, we found a putative boundary for silencing spread just below the shoot apical meristem that blocks movement upward from the root and hypocotyl. Finally, as PD aperture affects protein movement in embryos, we compared the movement patterns of protein tracers versus the movement patterns of RNA silencing signals. The data reveal that silencing signal complexes move to a similar extent as soluble proteins between 27–54 kDa.
Several lines of evidence support plasmodesmata (PD) as being the obligatory pathway through which RNA silencing signals move in plants (Voinnet and Baulcombe, 1997). PD are dynamic intercellular channels that interconnect cell-wall encased plant cells (Oparka, 2005). Silencing does not spread into guard cells that lose their PD connections during differentiation (Voinnet et al., 1998). Graft studies demonstrate the transmission of silencing from stock to scion via the vascular system; silencing in leaf cells surrounding the vasculature implies that silencing signals move there via PD connected to the vasculature (Palauqui et al., 1997). Both RNAs, such as viral RNAs, and proteins move between cells via PD, and such intercellular communication is likely to be essential to coordinate the programming of cell, tissue and organ formation (Oparka, 2005). However, no studies directly address whether PD aperture in different tissues can affect the extent of RNA silencing signal(s) movement.
There are two types of RNA silencing signal movement (Himber et al., 2003; Voinnet, 2005). Long-range movement mimics that observed during systemic infection by plant viruses, whereby silencing spreads from an initial leaf to additional leaves along the plant axis via the vascular system. Long-range silencing absolutely requires RDRPs, suggesting that propagation depends on re-amplification of the signal(s) (Himber et al., 2003; Voinnet, 2005). Short-range movement occurs, on average, only within the first 10–15 cells from the initiating silencing cells (which produce 21-nucleotide (nt) primary siRNAs), and is dependent on DICER-LIKE4 (DCL4), but is independent of RDRP activity (Dunoyer et al., 2005). Short-range movement can also be more extensive. This occurs by reiterated cell-to-cell movement in waves of 10–15 cells at a time. This extensive movement requires the production of secondary siRNAs, a process called transitivity, during which primary siRNAs (or miRNA) induce dsRNA synthesis through the activity of RDRP6 (also called SDE1), and subsequently produce secondary siRNAs that are not identical in sequence with the primary siRNAs. Interestingly, transitivity is only induced in response to exogenous sequences. Silencing of endogenous transcripts only induces short-range silencing (Dunoyer et al., 2005; Himber et al., 2003; Parizotto et al., 2004).
The Arabidopsis embryo is an advantageous and reliable system to observe spatial and temporal patterns of gene expression, and intercellular trafficking (Kim et al., 2002, 2005a; Stadler et al., 2005). In addition, dynamic changes of PD aperture and a regional map of symplastic subdomains were recently described for the Arabidopsis embryo (Kim et al., 2005b). Here, we set out to determine if (i) RNA silencing can be observed during embryogenesis, and if so, whether (ii) regions of the embryo that contain cells with distinct PD aperture exhibit different degrees of RNA silencing signal movement.
Results and discussion
Our approach is to induce expression of RNA silencing ‘triggers’ in specific cells of the embryo. First, to ensure accuracy in interpretation of expression patterns we chose to use the well-characterized DR5rev promoter that has been used extensively to monitor auxin-responsive gene expression both during embryogenesis and post-germination (Benkova et al., 2003; Friml et al., 2002; Friml et al., 2003; Sabatini et al., 1999; Xiao et al., 2006). For example, when this promoter is fused to sequences for green fluorescent protein (GFP) and an endoplasmic reticulum (ER) retention signal, it consistently marks cells responding to endogenous levels of the plant hormone auxin (see also below). Second, to ensure an efficient silencing phenotype and accuracy in its interpretation, we chose a target gene that has been previously used to demonstrate gene silencing in adult leaves, the ubiquitous and strongly expressed small subunit 3b of Rubisco (Rub-3b); silencing of Rub-3b leads to the loss of chlorophyll autofluorescence (Himber et al., 2003). The Rub-3b RNA silencing trigger is dsRNA (self-pairing inverted – repeated mRNA) (Waterhouse et al., 1998), containing partial sequence homology to the target mRNA (Figure 1a). There are three potential outcomes resulting from the induction of RNA silencing in embryos: no movement of silencing (cell-autonomous), extensive movement to the entire embryo body (complete silencing) or partial silencing in between the two extremes of no movement and complete movement. Figure 1h illustrates these outcomes for the silencing of Rub-3b in regions of the embryo that induce the silencing trigger by the DR5 promoter in response to endogenous auxin at the tips of the cotyledons and root.
Starting from the early torpedo stage, Arabidopsis wild-type embryos exhibit strong red autofluorescence of chlorophylls when observed under blue light (Figure 1e,f). Only root tips and provascular strands do not exhibit red autofluoresence, as cells located in these regions do not contain proplastids (Figure 1e,f, white arrows). Transgenic embryos transformed with the Rub-3b trigger construct exhibit spatially delimited disappearance of red-chlorophyll fluorescence, as manifested by the black regions at the tips of cotyledons and root (Figure 1b–d,g). This black pattern is consistent with the locations of cells where the DR5rev promoter is most active (Friml et al., 2003), i.e. at the tips of the cotyledons and root (Figure 2, Figure 3d, and represented as black regions in Figure 1h). We interpreted the loss of chlorophyll autofluorescence as a direct consequence of PTGS against Rub-3b mRNAs, as published previously in adult plants using a similar construct (Himber et al., 2003).
To demonstrate that the silencing of the Rub-3b gene is mediated by siRNAs of 21 nt, total RNA was purified from two different silencing lines at the young seedling stage. Line 2–3 exhibits a consistent silencing phenotype in seedlings manifested by the white circular areas at the tips of open cotyledons (Figure S4e). This photobleaching phenotype has been seen in post-germination tissues in a silenced line in which the expression of the silencing trigger in phloem cells leads to photobleaching of leaf veins (Himber et al., 2003). Seedlings from an independent line 2–5 show a weaker silencing phenotype with either no or smaller white areas (data not shown). Interestingly, the silencing phenotype between these two lines is correlated with the quantity of siRNA detected, i.e. we detected 21 siRNAs in line 2–3 but not in line 2–5. Thus, the visible extent of silencing qualitatively correlates with the level of siRNAs. However, as the silencing area was variable, even within one line, we are unable to make a more precise quantitative correlation. The detection of 21-nt siRNAs complementary to the Rub-3b gene in lanes corresponding to line 2–3 (Figure S4f) strongly suggests the silencing phenotype observed in seedlings (as well as in embryos) is mediated by siRNAs. We chose line 2–3 for further silencing characterization.
In line 2–3 embryos, in addition to consistent silencing at the tips of cotyledons and at the base of the root, where the trigger is expected to be highly expressed, we also observed extensive chlorophyll disappearance along the length of the hypocotyl and in the upper root (Figure 1b–d,g, Figure 2a). These results suggest that silencing signals have moved cell-to-cell from their site of synthesis at the root tip up into the hypocotyl. In contrast, the spread of silencing in the cotyledons was more limited to a few cells surrounding the region where the trigger dsRNA is expressed as based on DR5rev promoter activity evident in DR5rev ER-tethered GFP (ERGFP) embryos (Figures 2, 3d,f). These results suggest more restricted movement of silencing signals from cotyledon tips and provascular strands (Figure 2b). There is a dramatic difference in the level of silencing signal movement between the cotyledons and the hypocotyl. This result agrees with our recent results demonstrating that PD aperture is greater in hypocotyls compared with cotyledons (Kim et al., 2005b). Thus, silencing signal movement, like protein movement, is dependent on PD aperture. See Kim et al. (2005b) for an extensive analysis of protein tracer movement during Arabidopsis embryogenesis.
Interestingly, an apparent boundary between silenced and non-silenced regions can be visualized in the upper region of hypocotyls (Figure 1b–d yellow arrows, and Figure S1–S3). We determined the relative length of the hypocotyl-silenced zone relative to the overall lengths of embryos in late-heart, early- and mid-torpedo stages. The values obtained indicate that the relative distances travelled by silencing signals from the root were similar, between 42–50% of the embryo length, during three successive developmental stages (Figure 2c, Figure S1–S3). Thus, the apparent boundary, which restricts silencing spreading upward from the hypocotyl, consistently locates just below the shoot apical meristem (SAM). This boundary is easier to see in the earlier-stages embryos shown in Figure S1–S3.
Higher magnification images show that the reduction of red fluorescence along the hypocotyl is gradual, forming a gradient in intensity along its axis (Figure 1d and g). If we assume, as suggested previously by Dunoyer et al. (2005), that the efficacy of silencing (i.e. degradation of target mRNAs) is proportional to the number of RNA silencing signal(s) transported between cells from their initial site of synthesis, this gradient suggests that silencing signals move in a diffusive manner along the length of hypocotyls, analogous to the non-targeted movement of exogenous proteins such as GFP (Crawford and Zambryski, 2000; Crawford and Zambryski, 2001).
To further support that the silencing observed in embryos is mediated by PTGS, we developed a new technique, which consists of exogenous loading of dsRNA complementary to the target mRNA. In this experiment, transgenic embryos constitutively expressing ERGFP (line GFP142) were incubated with dsRNA GFP silencing trigger following a similar protocol described for our fluorescent tracer-loading assay (Kim et al., 2002). Basically, when embryos are released from their seed coats, breaks are introduced into the outer cell layer and wall. These breaks provide entry points for tracers that then move into embryos according to the aperture of PD; small tracers more readily move cell-to-cell than larger tracers (Kim et al., 2002). Using this approach we observed dramatic reduction in GFP fluorescence in embryos incubated in the presence of the GFP silencing trigger compared with control embryos (see Figure S4a,S4b). Performing such studies in embryos is technically challenging and has inherent variability because of the random generation of entry points for exogenously added tracers. Nevertheless, we consistently observed silencing only in the presence of dsRNA GFP in several experiments. Interestingly, the silencing of GFP by this method was not observed in older torpedo embryos. We previously found that PD aperture is down-regulated after the mid-torpedo stage for the uptake of 10-kDa fluorescent dextrans. These new data suggest that loading of dsRNA may be similarly restricted by either PD aperture or PD activity in later stages of embryogenesis.
Based on our previous work with GFP movement in embryos (Kim et al., 2005a,b), we suggest that the difference in extent of silencing signal movement between hypocotyl and cotyledons is likely to reflect differences in tissue-specific PD aperture/activity. To further estimate the extent of silencing signal movement, we compared the dimensions of the areas expressing ERGFP, which marks the site of silencing signal synthesis, to the areas of Rub-3b-silenced cells (Figure 2) in mid-torpedo embryos. The estimated distance (from 10 embryos) for silencing movement was 23 μm in cotyledons, and 250 μm in hypocotyls. Embryo cells range in size from 7 to 10 μm (Bowman, 1994). Thus, silencing signals move two–three cells in cotyledons and 25–35 cells in hypocotyls. The extensive movement in hypocotyls agrees with observations that PD are more dilated in this region compared with cotyledons (Kim et al., 2005a,b). These results furthermore imply that primary silencing signal movement in embryo hypocotyls (25–35 cells) is more extensive than observed in mature leaves (10–15 cells) (Himber et al., 2003).
Next, we compared the patterns of silencing signal movement to soluble protein movement. The molecular nature of the RNA silencing signal is unknown, but it surely contains RNA, and all available data suggests it moves as a ribonucleic acid protein complex (Yoo et al., 2004). Here we used the above DR5rev promoter to drive single- and double-sized soluble GFP expression. A control DR5rev-ERGFP line reveals where GFP is expressed as the ER retention signal sequesters GFP, inhibiting its intercellular movement. The pattern of GFP fluorescence in this ERGFP control line remains unchanged from early to late torpedo development (not shown), showing GFP fluorescence at root and cotyledon tips, and weakly in provascular strands (Figure 3d,f).
Comparison of the GFP signal in the lines expressing untethered soluble GFP then reveals the extent of free GFP cell-to-cell movement. GFP1X moves extensively throughout all cells and tissues of early torpedo embryos (Figure 3b). This indicates PD aperture size exclusion limit (SEL) is sufficient to allow the spread of a 27-kDa protein originating form the sites of DR5 promoter activity at the cotyledon and root tips, in agreement with our previous results (Kim et al., 2005a,b). In contrast GFP2X exhibits significantly restricted movement compared with GFP1X. GFP2X only shows limited movement from cotyledon tips and provascular traces at the early torpedo stage (Figure 3c,e). However, GFP2X does not move upwards from the root tip. This latter restriction is consistent with previous data where GFP2X expression was induced in embryonic root meristematic cells, and did not move upwards (Kim et al., 2005a,b). The new data here (GFP2X expression under the control of the DR5rev promoter) further confirm that PD at the root/hypocotyl junction form a boundary with a specific SEL allowing traffic of proteins between 27 and 54 kDa.
We assume that the silencing signal complex moves in a passive diffusive manner, as evident from the gradient in degree of silencing along the length of the hypocotyl (Figure 1g). Non-targeted protein movement also occurs in a diffusive manner that is dependent on protein size (Crawford and Zambryski, 2000; Kim et al., 2005a,b). Thus, we estimated the size of the putative silencing signal complex by comparing the area of gene silencing (Figure 3a) with the extent of GFP movement following expression under the same DR5 promoter (Figure 3b,c,e). GFP1X moves throughout the embryo; however, GFP2X does not move up into the hypocotyl. The putative silencing complex moves past this hypocotyl junction, but its movement is not as extensive as GFP1X. We extrapolate these data to suggest that silencing signals are likely to move as a complex with a size in between GFP1X and GFP2X. For simplicity, we have assumed that the rates of macromolecular cell-to-cell diffusive movement are proportional to their size (in daltons). However, when available, the effective hydrodynamic size of the silencing complex, measured by its Stokes radius, must also be taken into account as a measure of its effective size in relation to the PD SEL (Heinlein and Epel, 2004).
The above data imply that silencing signal movement is more restricted in cotyledons. To ensure that this restriction is not the result of an inability of cotyledons to generate silencing signals, we performed an additional experiment to test for endogenous miRNA-mediated RNA silencing in cotyledons. We used an available transgenic line constitutively expressing a modified version of ERGFP mRNA that contains a complementary sequence to miRNA171 in its 3′ non-coding region (GFP-miR171.1) (Figure 4b) (Parizotto et al., 2004).
Control late-torpedo embryos, constitutively expressing ERGFP under the control of the 35S promoter, showed GFP signals almost everywhere except at the lower region of the hypocotyl and the root (Figure 4a). In contrast, in late-torpedo embryos constitutively expressing the GFP-miR171.1 construct in the sde1 mutant background, ERGFP signals were observed in distinct cell types, most strongly in provascular strands and cells immediately surrounding these regions (Figure 4b). In the sde1 background, silencing signals cannot be amplified to permit cell-to-cell movement, and regions lacking GFP fluorescence suggest the presence of cells expressing endogenous miRNA171. The pattern of loss of GFP signal in cells surrounding the embryonic provascular system is reminiscent of the pattern of silencing observed in adult sde1 plants expressing GFP-miR171.1 (Parizotto et al., 2004). In contrast, in the presence of functional SDE1, transitivity and secondary siRNA production are induced, and these secondary siRNAs now can move into provascular strands and target GFP-miR171.1 transcripts. Consequently there is a complete absence of GFP signals in wild-type GFP-miR171.1 embryos (Figure 4e), as also reported for adult plants (Himber et al., 2003; Parizotto et al., 2004). Thus, silencing signals can be induced in cotyledons and move from non-vascular into vascular tissues, whereas movement in the reverse direction is more limited (Figure 2). We also observed miRNA-induced silencing in earlier, late-heart and mid-torpedo embryos (Figure S4c,d).
Here, we show that (i) PTGS occurs during Arabidopsis embryogenesis, (ii) different regions of the embryo are capable of generating silencing signals that move and (iii) the extent of silencing signal cell-to-cell movement is correlated with PD aperture in the cotyledons and hypocotyl. The data further suggest the regulation of primary silencing signal movement by a symplastic boundary close to the embryonic shoot meristem. The present analyses measure primary siRNA movement in response to a trigger homologous to an endogenous transcript Rub-3b; in contrast to previous analogous studies in adult leaves where Rub-3b siRNA moves 10–15 cells, cells in the embryonic hypocotyl allow Rub-3b siRNA movement of up to 35 cells. These results agree with numerous reports in the literature that have established that PD apertures are more dilated in immature versus mature tissues (reviewed in Kobayashi et al., 2005). Comparison of the extent of movement of silencing signals and GFP tracers indicates that the size of the silencing signal complex is between 27 and 54 kDa. In summary, these results support the general idea that the extent of both soluble protein and silencing signals short-range movement is governed by diffusion, and regulated by PD aperture.
We chose a binary vector pGSA1285 (http://www.chromdb.org) for expression of dsRNA siRNA triggers in transgenic embryos. The original vector pGSA1285 has a chloramphenicol resistance (CmR) gene for bacterial selection, a Kanamycin resistance (NPTII) gene for plant selection, a triple CaMV 35S promoter to drive the expression of the inverted repeat target sequence, and a 1352-bp uidA intron to stabilize the inverted repeat of the target gene fragment. We replaced the CaMV 35S promoter with an auxin-inducible promoter DR5rev, obtained by PCR amplification using the pS001 binary vector as a template (Friml et al., 2003) and primers s2-46up ‘GGAATTCGTCGACGGTATCGCAG’ and s2-46lo ‘TGTAATTGTAATTGTAAATAG’, previously cloned into the pTOPO cloning vector.
For PCR amplification of the Rub-3b sequence (Himber et al., 2003), we used a set of primers containing ‘inner’ restriction enzyme sites, either AscI or SwaI, and ‘outer’ restriction enzyme sites, either SpeI or BamHI, at the ends of each primer. A two-step cloning strategy, using the introduced restriction enzyme sites, assembles the inverted repeat in the binary vector pGSA1285. These same sites flank the uidA intron in pGSA1285. The final product contains an inverted repeat Rub-3b separated by the uidA intron.
The 3.3-kb STM promoter sequence from binary vectors containing the coding sequences of sGFPs either 1X or 2X (Kim et al., 2005b) was replaced by DR5rev promoter.
Arabidopsis growth and transformation
Seeds were stratified at 4°C for 2 days, before growth at 24°C with 16-h light /8-h dark cycles, either on soil or on plates with MS medium. Agrobacterium tumefaciens GV3101 carrying DNA constructs was used to transform Arabidopsis thaliana Col by the floral-dipping method.
From six T2 lines transformed with DR5rev sGFP1X (or 2X), two lines with strong and comparable expression of either 1X or 2XGFP were chosen for analyses. From seven T2 lines transformed with DR5rev Rub-buR showing a strong chlorophyll-silencing phenotype, line 2-3 had the most reproducible silencing phenotype and was chosen for analyses.
Seedlings and embryos were grown under greenhouse conditions and observed by epifluorescence and confocal laser scanning microscopy as described previously (Kim et al., 2002). Images were processed with either Adobe Photoshop (http://www.adobe.com) or Canvas 9 (http://www.deneba.com).
dsRNA GFP loading assay
dsRNA GFP was generated by in vitro transcription using T7 RNA-polymerase. The products were annealed by heating to 65°C and slowly cooling to 25°C (Saleh et al., 2006). RNAs were precipitated and resuspended in H2O. For each loading assay, either one or two siliques were opened and the released embryos were incubated in the presence of MS medium supplemented with 1% sucrose either with or without 12 μg of dsRNA GFP for 5 min. More MS medium then was added and the slides were incubated in a humid chamber at 25°C for 12–18 h. Three independent experiments were performed. One representative experiment consisted of 10 control and eight dsRNA-treated embryos (in total 18 embryos); all treated embryos demonstrated clear evidence of dsRNA-induced silencing. Images were taken on an Zeiss Axiophot epifluorescence microscope using GFP filters. Images were enhanced using Adobe Photoshop.
We thank Olivier Voinnet for providing the GFP-miR171.1 and GFP142 lines, Jiri Friml for providing the DR5rev ERGFP line and DR5rev plasmid DNA, Steve Ruzin and Denise Schieres from Biological Imaging Center for assistance and Howard Goodman for valuable discussion. We also thank Maria Carla Saleh for the gift of double-stranded RNA. Supported by NIH grant GM45244.