In higher plants a number of physiological processes are regulated by systemic RNA signaling molecules. This phloem-mediated remote-control system provides specific and efficient regulation to fine-tune many plant developmental programs. However, the molecular mechanism underlying long-distance movement of RNA remains to be elucidated. To this end, we examined the long-distance movement of GA-insensitive (GAI) RNA by Arabidopsis inflorescence grafting and RT-PCR analysis. Our results demonstrated that long-distance movement of RNA only occurred in specific transcripts. In addition, the sequences of GAI RNA are necessary and sufficient to target GREEN FLUORESCENT PROTEIN (GFP) RNA for long-distance movement, which indicates that the trafficking of GAI RNA is mediated by specific RNA motifs. Further analyses revealed that the motifs at coding sequences and 3′ untranslated regions of GAI RNA play important roles during RNA movement. In addition, the structure of the RNA rather than its specific sequence may also be important in GAI RNA trafficking. However, the secondary structure of GAI RNA is not the only factor to target RNA for long-distance movement, because recovery of the secondary structure of movement-defective GAI RNA only partially rescued RNA movement. Taken together, our results show that long-distance movement of non-cell autonomous RNA operates by specific RNA mobile elements.
By application of the RNA delivery system that has been established in the literature, a hypothetical model has been proposed to elucidate the molecular mechanism underlying long-distance movement of RNA in plants (Lucas et al., 2001). In this model, the cis-acting sequence elements that are located on non-cell autonomous RNA, termed ‘zip codes’, may interact with putative companion cell-specific zip code-binding proteins to form an RNA–protein complex. This RNA–protein complex is subsequently transported into sieve elements through plasmodesmata (Lucas et al., 2001). Consistent with this hypothesis, mutations altering the stem–loop structure of a viroid specifically disrupted entry of the viroid from the bundle sheath into mesophyll cells (Qi et al., 2004; Zhong et al., 2007 &Zhong et al., 2008). This evidence supports the idea that RNA structural motifs directly mediate the trafficking of viroid RNA. However, whether this structural motif is sufficient to target other cell-autonomous RNAs for long-distance movement has yet to be investigated.
In Arabidopsis and tomato, grafting experiments performed with transformant stocks of PCaMV35S–GAI and PCaMV35S–GFP showed that GAI but not GFP RNA can move long distances, which suggests that the trafficking of GAI RNA may be mediated by a motif that is absent from GFP RNA (Haywood et al., 2005). In contrast, efforts with a potato non-cell autonomous RNA, SUCROSE TRANSPORTER 1 (SUT1), failed to identify the cis-acting elements involved in SUT1 RNA movement (Kühn et al., 1997; Lalonde et al., 2003). Thus, whether the long-distance movement of plant endogenous, non-cell autonomous RNA is mediated by a specific RNA motif remains to be determined.
GAI encodes a nuclear protein that belongs to the DELLA subfamily of GRAS transcription factors (Pysh et al., 1999). Arabidopsis has five DELLA proteins: GAI, RGA, RGL1, RGL2 and RGL3. The function of these proteins is partially redundant in negative regulation of GA responses (Peng et al., 1997; Silverstone et al., 1998; Dill and Sun, 2001; Wen and Chang, 2002; Tyler et al., 2004). The application of GA stimulates the interaction between DELLA proteins and the GA receptor GIBBERELLIN INSENSITIVE DWARF1 (GID1), which leads to proteolysis or inactivation of the DELLA proteins and consequently relieves various GA-dependent responses (Olszewski et al., 2002; Schwechheimer, 2008). GAI may act in a non-cell autonomous manner. GAI RNA is detected in pumpkin phloem sap, and heterografting experiments have demonstrated that GAI RNA moves from the stock to the scion apex (Ruiz-Medrano et al., 1999; Haywood et al., 2005). In tomato grafting, the long-distance movement of DELLA-deleted GAI RNA is associated with phenotypic alterations in scion leaves, which suggests that GAI RNA is a systemic signaling molecule (Haywood et al., 2005).
In this study, we investigated whether specific RNA motifs are involved in long-distance trafficking of plant non-cell autonomous RNA. We found that the GAI RNA does indeed contain the specific RNA motifs that are necessary and sufficient to target GFP RNA, a cell-autonomous RNA, for long-distance movement. Deletion and linker-scanning analyses demonstrated that the motifs required for GAI RNA movement are located in the 3′ sequences of GAI RNA. Taken together, our data show that the RNA mobile elements in GAI RNA play important roles in RNA movement.
Long-distance movement of GAI RNA is impaired when ectopically expressed in mesophyll cells
To understand the molecular mechanism underlying GAI RNA long-distance trafficking, we first tested whether the movement of GAI RNA occurs only in specific cell types. The full-length GAI RNA, which contains 2146 nucleotides (nt) including 5′ (208 nt) and 3′ (340 nt) untranslated regions (UTRs), was amplified and driven by an Arabidopsis mesophyll-specific promoter, RbcS-2b (At5g38420; Kim et al., 2003), or a companion cell-specific promoter, SUC2 (Truernit and Sauer, 1995). To distinguish the transgene from the endogenous RNA, the nopaline synthase (NOS) terminator was included after the 3′ UTR of GAI, because it has been shown that this RNA fragment does not interfere the long-distance movement of GAI RNA (Haywood et al., 2005). For each construct, 10 independent transformants were selected as stocks and grafted with wild-type inflorescences. Two weeks after grafting, the scion apices were harvested for RNA extraction. In addition, the mature leaves of stocks were used as controls (Figure 1a). Reverse transcription-polymerase chain reaction (RT-PCR) was conducted to examine whether GAI RNA was transported from stocks to the scion apices. Total RNA of scion apices or stock leaves was used for RT reactions, and an aliquot of the first-strand cDNA was subjected to PCR with primers against GAI and NOS terminator (see Experimental Procedures). The PCR was conducted with 35 cycles, and the PCR products were visualized in agarose gels. For those samples that did not produce detectable signals in 35 cycles, the number of PCR cycles was increased to 40 cycles. The non-detectable sample was defined as no visualized PCR products in both 35 and 40 cycles. The PCR was conducted at least twice for each sample to make sure the data are representative. When wild-type scions were grafted onto PSUC2–GAI transformants, the transgenic GAI RNA was detected in 7 out of 10 wild-type scions (Figure 1b; represented as 70% of the scion detection rate), which suggests that the GAI RNA readily moved long distances when expressed in companion cells. In contrast, GAI RNA was detected in only 1 out of 10 scions grafted onto PRbcS2b–GAI transformant stocks (10% scion detection rate), which suggests that the movement of GAI RNA was limited when expressed in the mesophyll cells (Figure 1c). In control experiments, the detection rate of GAI RNA from the scions was 92% (46 out of 50 grafts) when wild-type scions were grafted onto PCaMV35S–GAI transformant stocks, or 0% (0 out of 40 grafts) when PCaMV35S–GFP transformants were used as stocks (Figure 1b; Haywood et al., 2005).
As it is possible that low-level accumulation of transgenic GAI RNA in PRbcS2b–GAI stocks may have accounted for the lack of detected GAI RNA in scions, RT-PCR was used to examine the level of transgenic GAI RNA in PRbcS2b–GAI and PSUC2–GAI transformants. The level of transgenic GAI RNA in stocks was not correlated with the detection of translocated GAI RNA in scions (Figure S1), which suggests that a high accumulation of GAI RNA in the stock is not sufficient to trigger RNA movement. Thus, the machinery required for GAI RNA trafficking might be mainly located in companion cells.
RNA long-distance movement occurs only with specific transcripts
In Arabidopsis, sequence comparison of the five DELLA proteins, GAI, RGA, RGL1, RGL2 and RGL3, revealed that the coding regions of GAI paralogs are conserved, but the UTRs are relatively diverse (Figure S2). To test whether the sequence variations among GAI paralogs may affect RNA trafficking, we amplified the full-length cDNA (containing 5′ and 3′ UTRs) of the five GAI paralogs and driven by a CaMV35S promoter. Again, Arabidopsis inflorescence grafting and RT-PCR analysis were performed to examine the long-distance trafficking of transgenic RNA. The detection rate of the GAI RNA from wild-type scions grafted onto PCaMV35S–GAI stocks was 90% (Table 1). In contrast, the detection rate of RGA, RGL1, RGL2 or RGL3 RNA was greatly reduced from the scions that had been grafted onto PCaMV35S–RGA, –RGL1, –RGL2 or –RGL3 transformant stocks (Table 1). Therefore, although the five GAI paralogs share high sequence similarity, only GAI RNA efficiently moves long distances.
Table 1. Arabidopsis grafting performed with wild-type scions grafted onto individual GAI paralog transformants
Scion detection (%)
No. of grafts
No. of lines
GAI RNA is sufficient to target GFP RNA for long-distance trafficking
If GAI RNA contains the motif required for RNA long-distance movement, this motif may be able to target other cell-autonomous RNAs to move over long distances. To test this possibility, Arabidopsis transformants carrying PCaMV35S–GAI–GFP or PCaMV35S–GFP–GAI transgenes were used as stocks and grafted with wild-type scions. In control experiments, GFP RNA was not detected in wild-type scions with PCaMV35S–GFP used as a stock (Figure 2; Haywood et al., 2005). However, when PCaMV35S–GAI–GFP or PCaMV35S–GFP–GAI transformants were used as stocks, the RNA of GFP–GAI and GAI–GFP was detected in wild-type scions (Figure 2a). In our 80 grafting samples, detection of translocated RNA was highly reproducible (Figure 2b). Thus, the information contained within GAI RNA is sufficient to target cell-autonomous RNA for long-distance delivery.
Deletion analyses were conducted to locate the potential motifs required for GAI RNA movement. To test whether the 5′ or 3′ UTR is required for GAI RNA trafficking, GAI with different truncated UTRs was generated, and driven by a CaMV35S promoter. The NOS-terminator was again used to distinguish the transgene from the endogenous GAI. Arabidopsis transformants carrying GAI transgenes were first analyzed by RT-PCR with specific primers against the transgene, and the transformants with similar RNA accumulation level of transgene were used as stocks (data not shown). Our grafting results indicated that the 5′ UTR of GAI RNA was dispensable for RNA trafficking (Figure 3a,c). In contrast, deletion of the GAI 3′ UTR greatly reduced the rate of detection of GAI RNA in the scions, which suggests that the 3′ UTR is required for long-distance movement of GAI RNA (Figure 3a,c). To examine whether the 3′ UTR of GAI RNA is sufficient to target GFP RNA movement, GFP was fused with the GAI 3′ UTR and driven by a CaMV35S promoter. Chimeric GFP RNA was not detected in the wild-type scions grafted onto transformants carrying GFP–GAI1808–2146 (3′ UTR), GFP–GAI1752–2146 (50 nt of coding sequence +3′ UTR) or GFP–GAI1700–2146 (100 nt of coding sequence +3′ UTR) transgene, which indicates that the 3′ UTR of GAI RNA is not sufficient for RNA movement (Figure 3b,d). However, chimeric RNA was detected in wild-type scions grafted onto transformants carrying PCaMV35S–GFP–GAI995–2146 transgene, which indicates that the RNA of GAI995–2146 is sufficient to target GFP RNA for long-distance movement (Figure 3b,d). Thus, the GAI995–2146 RNA fragment is necessary and sufficient for long-distance trafficking of RNA. In addition, our data also suggest that at least two RNA mobile elements, one located in the GAI coding sequence and the other in the 3′ UTR, are required for movement of GAI RNA. Alternatively, the RNA structure created by folding of the GAI coding sequence and the 3′ UTR may determine the long-distance trafficking of GAI RNA.
Linker-scanning analysis of the long-distance movement of GAI995–2146 RNA
Linker-scanning analysis was employed to further characterize the RNA mobile elements within GAI995–2146 RNA. Site-directed mutagenesis was used to create 26 sequence-replaced mutants (M1–M26) across the 995–2146 nt of GAI RNA (Figure S3). Arabidopsis transformants expressing mutant GAI RNA were used as stocks and grafted with wild-type scions. The scion detection rate of the GAI995–2146 RNA was 67% when wild-type scions were grafted onto PCaMV35S–GAI995–2146 transformant stocks (Figure 4a,c; Table S1). However, the scion detection rate of mutated GAI RNA was reduced to 30% or less when M1, M3, M4, M5, M7, M11, M13, M14, M17 and M18 transformants were used as stocks, which suggests that these regions are important during GAI RNA movement (Figure 4c, Table S1). When grafting used M6 and M9 stocks, the scion detection rate of the GAI RNA was reduced to 44 and 50%, respectively. In contrast, the scion detection rate of mutated GAI RNA was similar to or slightly better than that of GAI995–2146 when M2, M8, M10, M13, M15, M16 and M19–M26 transformants were used as stocks (Figure 4c, Table S1). The mutations leading to defective trafficking of GAI RNA were mainly located in three regions (motifs A, B and C), which suggest that multiple motifs are required for movement of GAI RNA (Figure 4c). In addition, because two-thirds of the mutations occurring in the fragment of GAI995–2146 strongly or partially disrupted RNA movement, the RNA structure may play a more important role in targeting GAI RNA for long-distance movement in RNA trafficking.
In yeast and animal cells, it has been shown that a specific stem–loop structure is required to target RNA for asymmetric localization (Chartrand et al., 1999; Gonzalez et al., 1999; Kloc et al., 2002). To test whether the stem–loop structure of GAI RNA is involved in the movement of RNA, we predicted the secondary structure of GAI995–2146 RNA using the computer program mfold (Zuker, 2003). We selected the M4 mutation for further analysis, because the sequence of M4 was located on a separate stem–loop arm and mutations occurring on this arm (M4 and M5) strongly reduced RNA movement to 20 or 10% (Figure 5). We speculated that if a specific stem–loop structure participates in GAI RNA movement, the recovery of this structure might rescue RNA movement to the wild-type level. To this end, the nucleotides that complemented the M4 sequence were mutated to recover a stem–loop structure similar to that of the wild type (Figure S4). Computer prediction showed that the entire structure of M4-rev RNA was the same as GAI995–2146 RNA (data not shown). In 50 grafting samples, the scion detection rate of M4-rev RNA was increased to 43% (Figure 4b,c), which indicates that the recovery of the stem–loop structure only partially rescued the long-distance movement of GAI RNA. Thus, the stem–loop structure of GAI RNA is not the only factor determining RNA movement.
In this study we have shown that the long-distance trafficking of GAI RNA is mediated by specific RNA motifs. Based on deletion analyses, we have also shown that the GAI995–2146 RNA fragment is necessary and sufficient to target other cell autonomous RNAs for long-distance trafficking (Figure 3). In addition, our linker-scanning analysis revealed that at least three motifs are required for long-distance movement of GAI RNA (Figure 4a). Because the successful trafficking of GAI RNA from stock to scion apex requires multiple translocation steps, which include vascular entry, transport along the phloem and exit (Lucas et al., 2001), it is possible that each step may require a specific motif to interact with putative translocation machinery. Consistent with this notion, the results of our grafting experiments showed that the RNA movement is not associated with over-accumulation of GAI RNA in the stocks (Figures 1 and S1). Thus, it is possible that an active transport system, which involves RNA-binding proteins or other yet to be identified proteins, participates in the selection of specific transcripts for long-distance trafficking. Interestingly, the long-distance movement of GAI RNA was significantly higher when GAI was expressed in the companion cells rather than in the mesophyll cells (Table 1). This suggests that the machinery required for GAI RNA movement is mainly located in companion cells. However, because the cellular boundary may serve as a regulatory point for viroid RNA trafficking (Qi et al., 2004), it is possible that GAI RNA moves efficiently from cell to cell in mesophyll cells but in a limited fashion across the boundary into companion cells.
Our linker-scanning analysis showed that over two-thirds of mutations occurring at GAI995–2146 fragment affect GAI RNA movement, which suggests that the structure rather than specific sequences is important for recognition by RNA movement machinery. The recovery of the stem–loop structure in movement-defective GAI RNA only resulted in partial rescue of RNA movement (Figure 4b,c). Therefore, the stem–loop structure may not be the only factor for the recognition of RNA movement machinery. More detailed analyses are required to identify the nature of the RNA mobile elements. In viroids, it has been shown that the tertiary structure of the RNA motif plays an important role during the vascular entry of viroid RNA (Zhong et al., 2007). It is possible that the tertiary structure of GAI RNA or other factors may also contribute to the interaction with trans-acting factors. Interestingly, RNA tertiary structures are also important for asymmetric localization of yeast ASH1 mRNA (Chartrand et al., 1999; Gonzalez et al., 1999). This may raise the intriguing question of whether the mechanism of RNA long-distance movement and RNA asymmetrical localization shares the same evolutionary origin.
The scion detection rate of mutant GAI RNA was strongly affected when mutations occurred in motif C (M17 and M18), indicating that motif C is important for RNA movement (Figures 4 and S3). However, the insertion of GFP into the junction between the coding sequence and the 3′ UTR did not disrupt the movement of chimeric GAI–GFP RNA (Figure 2), suggesting that motif C may act separately from other motifs during GAI RNA trafficking. Alternatively, it is possible that specific sequences, or RNA structures of individual motifs, may interact with putative RNA movement machinery and assemble into a complicated RNA–protein complex. In yeast, it has been demonstrated that the interaction between RNA structure and a protein complex is important for RNA asymmetric localization (Chartrand et al., 1999, 2001). Whether a similar model can be applied to GAI RNA long-distance movement needs further investigation.
The primary sequence comparison among five GAI paralogs indicated that RGA is more similar to GAI (Figure S2). However, our grafting analyses showed that the movement efficiency of RGL3 is slightly better than RGA, RGL1 or RGL2 (Table 1). This result is consistent with the notion that the structure rather than the primary sequences of GAI RNA participates in RNA movement. Whether the structure of RGL3 is more accessible to RNA movement machinery remains to be investigated.
When GAI 3′ UTRs were deleted, the detection of translocated GAI RNA from the scions was significantly reduced (Figure 3a). The truncated or mutant forms of GAI RNA may have been rapidly degraded before they could interact with the RNA trafficking machinery. However, our RT-PCR analyses showed that accumulation of the UTR-truncated or mutated GAI RNA in most of the GAI transformants was not significantly reduced (data not shown). Therefore, RNA stability was unlikely to be a major factor affecting the movement of GAI RNA in our experiments.
In higher plants, several thousand RNAs have been identified from analyses of the phloem sap RNA composition (Sasaki et al., 1998; Ruiz-Medrano et al., 1999; Lough and Lucas, 2006; Kehr and Buhtz, 2008). Because the sieve elements are enucleate cells, these phloem sap RNAs might be synthesized in the companion cells (or other cell types) and then transported into the sieve element. This finding raises the possibility that most, if not all, of the phloem sap RNA may have the ability to move from companion cells into sieve elements. Currently, it has been shown that at least four mRNAs (BEL5, LeT6, KNOTTED1 and GAI) are able to move for long distances or cell-to-cell in Arabidopsis or tomato (Lucas et al., 1995; Kim et al., 2001, 2005; Haywood et al., 2005; Banerjee et al., 2006). However, the sequence comparison or secondary structure prediction between GAI and BEL5 or other phloem sap RNAs failed to identify the conserved domain among these genes (data not shown). It is possible that more than one RNA trafficking machine participates in the movement of individual non-cell autonomous RNA. Consistent with this hypothesis, proteomic or mass spectrometric analysis of pumpkin or melon phloem sap proteins has identified a set of putative RNA-binding proteins, including a polypyrimidine tract-binding protein, RBP50 (Gómez et al., 2005;Lin et al., 2009; Ham et al., 2009). These proteins may be involved in the translocation of phloem-mobile RNA.
Arabidopsis thaliana plants were obtained from the Arabidopsis Biological Resource Center (ABRC, Columbus, OH, USA) and grown in growth chambers under 16-h/8-h, 22°C/20°C day/night cycle, under white fluorescent light at an intensity of 100 μmol m−2 sec−1.
The full-length GAI cDNA was described previously (Haywood et al., 2005). The Arabidopsis full-length RGA, RGL1, RGL2 and RGL3 cDNAs were amplified by RT-PCR with RNA extracted from Landsberg erecta (Ler) seedlings and gene-specific primers (sequences of the primers are in Table S2). For generation of GAI, RGA, RGL1, RGL2 and RGL3ΔDELLA lines, the full-length cDNA served as a template to generate a 51-bp deletion at the DELLA domain, and GAIΔDELLA was used to generate all the other GAI deletion lines by PCR amplification. To generate GFP–GAI and GAI–GFP clones, the GFP fragment was directly cloned into GAI full-length cDNA, which introduced a BamHI site by site-directed mutagenesis at positions 209 nt or 1807 nt of GAI, respectively. Polymerase chain reaction with suitable primers (Table S2) was used to create the varieties of GAI deletion clones. All the constructs were confirmed by sequencing and subcloned to binary vectors containing a CaMV35S or a SUC2 promoter. Arabidopsis was transformed by floral dip transformation (Clough and Bent, 1998). The transformants were selected on MS medium containing 40 μg ml−1 hygromycin. The antibiotic-resistant plants were transferred to soil for further analysis.
Arabidopsis inflorescence grafting
Arabidopsis grafting experiments were performed using the following procedures: Arabidopsis plants (Ler ecotype) were grown for 4–5 weeks until bolting. The inflorescence (about 4 cm long, containing two to three lateral floral buds) from scions was cut into a V-shape under microscopy and inserted into the stock, which had a vertical cut in the middle of the primary inflorescence and a piece of polyethylene tubing was inserted over the base of the cut bolt. The grafting junction was secured with the polyethylene tubing and sealed with Parafilm. Grafted plants were kept at high humidity for 7 days then returned to normal growth conditions for another week. Scion RNA samples were extracted 14 days after grafting.
RNA extraction and RT-PCR analysis
Total RNA from plant tissues was extracted using TRIzol reagent (Invitrogen, http://www.invitrogen.com/). For RT-PCR analysis, 5 μg of total RNA was used in RT reactions performed with oligo(dT)20 and SuperScript III reverse transcriptase (Invitrogen). One microliter of cDNA was used for the PCR reaction with the following conditions: 1 min at 94°C for 1 cycle; 30 sec at 94°C, 30 sec at 60°C, 1 min at 68°C for 35 cycles and 10 min at 68°C for 1 cycle. Specific primers against the transgene and NOS terminator were used for PCR reactions to distinguish transgenes from endogenous genes (primer sequences are in Table S3). To minimize the bias during PCR, the same primers were used for each set of experiments. The PCR was conducted at least twice for each sample to make sure the data were representative. An aliquot (5 μl) of PCR products was separated on 1.5% agarose gels.
To generate the linker-scanning mutations on GAI RNA, the plasmid containing the GAI995–2146 fragment served as a template for site-directed mutagenesis. These primers are listed in Table S4. Site-directed mutagenesis was performed under the following conditions: 1 min at 95°C for 1 cycle; 30 sec at 95°C, 1 min at 55°C, 8 min 30 sec at 68°C for 18 cycles and 1 min at 94°C, 1 min at 55°C, 10 min at 72°C for 1 cycle. The PCR product was digested with DpnI at 37°C for 1 h to remove the template DNA. To generate M4-rev, the M4-containing plasmid was used as a template for site-directed mutagenesis. All of the mutant clones were confirmed by sequencing analysis. The mutant constructs were driven by a CaMV35S promoter. For grafting experiments, five independent transformant plants from each mutant construct were used as the stocks and were grafted with wild-type scions.
We thank the ABRC for providing Arabidopsis seeds, and Drs Hsou-min Li and Shu-Hsing Wu and our lab members for critically reading the manuscript. This work was supported by grants NSC 95-2311-B-001-060 and NSC 96-2311-B-001-021-MY3 from the National Science Council, Taiwan.