Visualizing the spatiotemporal dynamics of native RNA molecules with single-transcript-level sensitivity in living plant cells remains a formidable challenge due to the high background noise caused by the unique cellular architecture of plants. Here, using transiently transformed Nicotiana benthamiana cells and stable Arabidopsis thaliana lines, we present a high-signal-to-noise RNA imaging platform that couples CRISPR–dCas13X with a binding-associated fluorescence stabilization strategy. We identified a novel plant-specific degron that selectively targets unbound probes for proteasomal degradation while preserving target-associated signal in a manner consistent with binding-dependent protection, enabling high-fidelity detection of endogenous mRNA puncta with minimal perturbation. Leveraging this platform, we characterized dynamics of FT mRNA trafficking in living plant cells, including plasmodesmata-targeting, cell-to-cell trafficking, and long-distance movement. Loss-of-function analysis further identified the established RNA-binding protein GRP7 as a factor that contributes to efficient FT mRNA transport. This work establishes a useful toolkit for investigating plant RNA biology and provides mechanistic insight into the systemic transport of florigenic signals.

Rational design of a GFP–cpScFv reporter for low-background RNA imaging

The CRISPR/dCas13 family-based SunTag systems have proven effective for endogenous RNA imaging in mammalian cells [17,21,22]. However, transferring these technologies to plant cells presents a unique physical barrier: the large central vacuole compresses the cytoplasm into a thin peripheral layer, which amplifies background noise by concentrating the pool of unbound scFv–GFP. This limitation, together with chloroplast autofluorescence, precludes high-contrast imaging (Fig 1A and 1B).

To overcome this, we re-engineered the fluorescence module to function as a “binding-activated” switch. We reasoned that converting scFv–GFP from a constitutively stable reporter into a conditionally stable one would selectively deplete the unbound cytosolic pool. To this end, we designed a circularly permuted scFv (cpScFv) by relocating the N- and C-termini near the antigen-binding pocket, enabling the fusion of a degron proximal to the complementarity-determining region (CDR) (Fig 1C). Mechanistically, this cpScFv–degron module was designed to function as a binding-associated stabilization switch for proteasomal turnover. In the unbound state, the exposed degron promotes degradation of free GFP–cpScFv, thereby reducing diffuse background. Upon binding to the SunTag array, the bound state is proposed to protect the degron from degradation machinery, stabilizing the fluorophore at target-associated sites. Thus, RNA-dependent recruitment converts molecular recognition into selective signal retention, enabling high-contrast imaging of true RNA loci (Fig 1D).

A proteome-wide screen identifies a plant-specific degron

Identifying a degron compatible with the plant cytosolic environment is critical. We first evaluated the efficiency of 10 previously reported degrons in Nicotiana benthamiana using a dual-fluorescence reporter system (Fig 2A). However, all tested degrons failed to sufficiently eliminate the GFP signal, with the lowest GFP/mCherry ratio remaining above 0.39 (Fig 2B and 2C), indicating insufficient instability for our purpose. Consequently, we performed a proteome-wide screen for endogenous instability domains. Using quantitative proteomics (LC–MS/MS) following MG132 treatment, we identified 86 proteins significantly upregulated upon proteasome inhibition and selected the top 30 candidates (Fig 2D). Screening the C-terminal peptides (CPs, 30 aa) of these candidates revealed that six peptides significantly induced GFP degradation (Figs 2E and S1). Among them, CP13—derived from a Nicotiana tabacum proteinase inhibitor I-B-like protein—exhibited the highest efficiency, suppressing the normalized GFP/mCherry ratio to <0.1 (Fig 2F). Truncation mapping further identified the C-terminal 21 amino acids as the core motif responsible for degradation (Fig 2G–2I). We designated this potent plant-specific degron as “cPIL” (C-terminal Proteinase Inhibitor-Like degron).

cPIL selectively depletes free reporters and enhances signal-to-noise in plant cells

To develop a SunTag-stabilized GFP–scFv reporter, we designed 10 circularly permuted GFP–scFv mutants with N- and C-termini near the antigen-binding pocket (Fig 3A). Imaging of tobacco epidermal cells co-expressing GFP–scFv mutants and SunTag revealed that only one mutant, with permuted termini adjacent to Asn194, produced fluorescent puncta when co-expressed with SunTag (Figs 3B and S2A). This mutant was designated “cpScFv”. The average number of puncta produced by GFP–cpScFv was comparable to that produced by wild-type GFP–scFv in SunTag-expressing cells (S2B Fig). These results indicate that cpScFv and wild-type scFv have similar SunTag-binding abilities.

We next examined whether the cPIL degron selectively destabilizes free GFP–cpScFv molecules in plant cells. Transient expression of either GFP–cpScFv or GFP–cpScFv–cPIL under the UBQ10 promoter in the same Nicotiana benthamiana leaves revealed strong diffuse nuclear and cytoplasmic fluorescence for GFP–cpScFv and GFP control, whereas the cPIL fusion showed substantially reduced bulk signal, similar to the negative control (Fig 3C and 3D). Quantification data showed that cPIL caused a greater than 3.34-fold decrease in fluorescence intensity of GFP (Fig 3E). Time-lapse imaging showed an 80% decrease in fluorescence intensity of GFP–cpScFv–cPIL after 30 min of cycloheximide treatment, while the decrease in fluorescence in GFP–cpScFv-expressing cells was significantly attenuated over the same period (Fig 3F and 3G). Immunoblot analysis revealed that cPIL significantly reduced GFP-fusion protein accumulation compared with the cPIL-free control during cycloheximide chase (Fig 3H). These results indicate that cPIL promotes the rapid degradation of GFP–cpScFv. Furthermore, treatment with the proteasome inhibitor MG132 supported proteasome-dependent degradation of GFP–cpScFv–cPIL (Fig 3I).

To determine whether SunTag binding protects GFP–cpScFv–cPIL at RNA foci, we co-expressed full-length dCas13X–24× SunTag, a gRNA targeting the mCherry reporter mRNA, and either GFP–cpScFv or GFP–cpScFv–cPIL in a stable mCherry-expressing line. In both configurations, we detected discrete nuclear and cytoplasmic puncta corresponding to RNA–dCas13X–SunTag complexes (Fig 3J). Quantitative image analysis showed that the cPIL fusion reduced diffuse background fluorescence by approximately 6.71-fold relative to GFP–cpScFv, whereas the mean intensity of individual foci remained at a similar level (Fig 3K and 3L). Consequently, the SNR increased by approximately 6.67-fold in the GFP–cpScFv–cPIL system (Fig 3M). These observations support a model in which free GFP–cpScFv–cPIL molecules are rapidly removed by the proteasome, whereas SunTag binding stabilized GFP–cpScFv–cPIL.

High-fidelity detection of endogenous ACT2 mRNA puncta

We asked whether the GFP–cpScFv–cPIL reporter combined with the dCas13X–SunTag system was capable of imaging endogenous mRNAs in plant tissues. The cytoskeletal ACT2 transcript was selected as the target RNA. A single plasmid containing the UBQ10::dCas13X–SunTag–IRES–GFP–cpScFv–cPIL–U6::gRNA construct was introduced into wild-type Arabidopsis (Col-0), and three independent transgenic lines were used for imaging (Fig 4A). We observed discrete nuclear and cytoplasmic puncta in both leaf and root cells (Fig 4B).

To confirm the single-transcript-level sensitivity, we benchmarked the GFP–cpScFv–cPIL system against smFISH in fixed leaves from the transgenic lines. We observed strong colocalization between GFP–cpScFv–cPIL puncta and smFISH puncta (Fig 4C and 4D), with an overall colocalization rate of 87.0 ± 3.8% (Fig 4E). GFP–cpScFv–cPIL showed a punctum-area distribution (0.1 μm² to 0.23 μm², consistent with the theoretical resolution limit of single-molecule RNA) and numbers of puncta per cell similar to those detected by smFISH (Fig 4F and 4G), while GFP–cpScFv–cPIL exhibited higher SNR than smFISH (Fig 4H), indicating improved imaging contrast for detection of endogenous RNA puncta in this assay. To further evaluate the nature of these puncta, we quantified the integrated intensity distributions of individual ACT2 dCas13X–SunTag puncta and corresponding ACT2 smFISH puncta. The ACT2 dCas13X–SunTag puncta were dominated by a major low-intensity population with a progressively decreasing higher-intensity tail, and this overall distribution pattern was broadly similar to that observed for ACT2 smFISH puncta (Fig 4I). Together, these data support the interpretation that most detected puncta represent single-transcript events under our assay conditions.

Furthermore, the specificity of GFP–cpScFv–cPIL for RNA imaging was validated using an MS2/MCP reporter system, in which a 24× MS2 tag was inserted into the 3′ UTR of ACT2 RNA to recruit mCherry-MCP. In tobacco epidermal cells co-expressing ACT2–24× MS2 reporter RNA, MCP–mCherry, GFP–cpScFv–cPIL, dCas13X–SunTag, and gRNA1, 85.8 ± 4.9% of GFP–cpScFv–cPIL puncta colocalized with MCP–mCherry foci, while a non-targeting gRNA markedly reduced the colocalization between dCas13X–SunTag puncta and MCP–mCherry foci (Fig 4J and 4K), indicating that signal detection is guide dependent. In addition, two independent ACT2-targeting gRNAs produced similarly high colocalization with the MCP–mCherry reporter (Figs 4L and S3), supporting reproducible detection across independent guides. Together, these results strengthen the specificity of the system.

Next, we assessed whether the GFP–cpScFv–cPIL system perturbed the target RNA. Measurements of ACT2 transcript abundance and protein levels in transgenic Arabidopsis thaliana carrying either the reporter system or the empty-vector control revealed that our system had no significant effect on target-gene transcript or protein abundance (S4A and S4B Fig). The RNA decay curve after treatment with actinomycin D indicated that the reporter system had no significant effect on the turnover of the target mRNA (S4C Fig). Collectively, these data support GFP–cpScFv–cPIL as an effective platform for RNA imaging with single-transcript-level sensitivity in plant cells.

Validation of generalizability across diverse endogenous targets

To evaluate the broad applicability of the dCas13X–SunTag/GFP–cpScFv–cPIL platform to diverse classes of endogenous plant mRNAs, we utilized the Arabidopsis root as an in vivo model to image endogenous mRNAs with distinct spatiotemporal profiles. We selected CORTEX4, IRX9, and ATPP2-A1—genes exhibiting cell-type–biased expression in the cortex, xylem, and phloem, respectively—as representative targets. For each gene, guide RNAs (gRNAs) targeting accessible regions within the coding sequence (CDS) were co-expressed with the optimized imaging module (Fig 5A–5C).

Confocal microscopy revealed discrete, diffraction-limited fluorescent foci for all three targets, which were clearly resolvable at the single-cell level. The spatial distribution of these foci was highly specific and concordant with established transcriptomic atlases [23–25]. CORTEX4 signals were enriched in cortical cell files, while IRX9 and ATPP2-A1 signals were strictly localized to the vascular domains (Fig 5A–5C). Quantitative analysis performed within the relevant target cell type for each marker confirmed that the platform captures expression dynamics across developmental stages (3, 5, and 7 days after germination [DAG]) (Fig 5D). Specifically, quantification was carried out in xylem cells for IRX9, phloem-domain cells for ATPP2-A1, and cortical cells for CORTEX4. Collectively, these results demonstrate that the platform enables high-contrast, cell-type–specific visualization of endogenous transcripts in intact plant tissues.

Visualizing the long-distance trafficking of FT mRNA

Finally, we applied our platform to further examine the dynamics of FLOWERING LOCUS T (FT) mRNA trafficking in living plant cells. We generated two types of transgenic lines expressing dCas13X–SunTag/GFP–cpScFv–cPIL with and without gRNA in wild-type backgrounds. We observed obvious colocalization of FT mRNA puncta with the PDCB1-BFP-indicated plasmodesmata (Fig 6A). Furthermore, time-lapse imaging showed rapid translocation between cells after docking at plasmodesmata-associated regions (Figs 6B and S5; S1 and S2 Videos). Quantitative analysis of FT mRNA punctum dynamics during cell-to-cell trafficking revealed that FT puncta typically first accumulated at plasmodesmata-associated regions, where they exhibited a relatively long mean dwell time of approximately 20 s before transfer (Fig 6C; S1 and S2 Videos). This residence phase was followed by a rapid intercellular passage step, with an average transit time of approximately 6 s across the cell boundary (Fig 6D; S1 and S2 Videos). Quantification of trafficking frequency further revealed an average of approximately 12 successful transfer events per minute per cell interface (Fig 6E). Together, these kinetic measurements suggest that FT mRNA trafficking involves at least two distinguishable phases: an initial residence phase at plasmodesmata-associated regions followed by rapid intercellular passage.

To further test long-distance transport, we generated two types of transgenic lines expressing dCas13X–SunTag/GFP–cpScFv–cPIL and either an FT-targeting or a non-targeting gRNA in transcript-null ft mutant Arabidopsis plants (ft_TG and ft_NC lines), which were grafted onto WT rootstocks (Fig 6F). We observed no fluorescent puncta in either ft_TG or ft_NC scions before grafting, whereas fluorescent puncta emerged in ft_TG scions but remained absent from ft_NC scions two weeks after grafting (Fig 6G), supporting the long-distance movement of FT mRNA from the WT rootstock to the ft mutant scion. The presence of full-length FT mRNA in the scion was confirmed by RT-PCR and sequencing (S6A and S6B Fig). Together, these data provide visual evidence supporting rootstock-to-scion FT mRNA transport.

GRP7 functionally contributes to efficient FT mRNA transport

To explore factors that may contribute to FT mRNA trafficking, we performed RNA pull-down assays using in vitro-transcribed, biotin-labeled full-length FT sense RNA, with antisense RNA as a control, followed by LC–MS/MS analysis of proteins enriched in the FT RNA pull-down (Fig 7A and 7B). From the differentially enriched candidates, we focused on three FT RNA-associated proteins: GLYCINE-RICH PROTEIN 7 (GRP7, AT2G21660), FRIGIDA-ESSENTIAL 1 (FES1, AT2G33835), and SWELLMAP-LIKE (SMPL, AT3G45950) (Fig 7C). Immunoblot analysis after RNA pull-down further confirmed enrichment of these candidate proteins in the FT RNA pull-down relative to the antisense RNA control (Fig 7D).

To assess their contributions to long-distance FT mRNA transport, we grafted the ft_TG line onto grp7, fes1, and smpl mutant rootstocks, respectively (Figs 7E and S7; S8 Table). FT-associated signals in ft_TG scions were strongly reduced when grp7 rootstocks were used, supporting a contribution of GRP7 to efficient rootstock-to-scion FT mRNA transport.

To further illustrate the role of GRP7 in FT trafficking, we next examined FT trafficking dynamics in the grp7 background. In grp7, FT-associated puncta still reached PDCB1-marked plasmodesmata-associated regions but remained at these regions, and no transfer event was observed during the 3-min imaging window (Fig 7F; S3 Video). These observations are consistent with a contribution of GRP7 to the intercellular transfer step following the association of FT puncta with plasmodesmata-associated regions.

Together, these results support a functional contribution of GRP7 to efficient FT mRNA transport, potentially during or after the association of FT puncta with plasmodesmata-associated regions.

Plant materials and growth conditions

Nicotiana benthamiana and Arabidopsis thaliana (Col-0) were grown in soil in a controlled greenhouse at 22 °C under long-day conditions (16 h light/8 h dark, light intensity ~120 µmol m⁻² s⁻¹). The A. thaliana mutant ft-10 (CS9869) was obtained from the Arabidopsis Biological Resource Center (ABRC). For plate-based assays, seeds were surface-sterilized and stratified at 4 °C for 2 d before sowing on half-strength Murashige and Skoog (1/2 MS) medium supplemented with 1% (w/v) sucrose and 0.8% (w/v) agar.

Plasmid construction and Agrobacterium strain

All DNA fragments encoding the circularly permuted scFv mutants (mscFv) and candidate degron peptides were synthesized by Tsingke Biotechnology Co. (Beijing, China). All vectors were based on pCAMBIA1300 derivatives. For the pUBQ10::mscFv constructs, the mscFv fragments were PCR-amplified and cloned downstream of the pUBQ10 promoter between the HindIII and NcoI restriction sites (S1 Table). For the pUBQ10::mCherry-P2A-GFP-CPs constructs, the degron peptide fragments were PCR-amplified and cloned into the C terminus of the mCherry–P2A–GFP cassette using homology-based recombination between the BamHI and BsrGI restriction sites. For the 10 degron peptides reported in previous studies, the corresponding oligonucleotides were synthesized (Tsingke), annealed at 95 °C for 5 min, and ligated into the same BamHI and BsrGI restriction sites at the GFP C terminus using T4 DNA ligase. To reduce fluorescence background, DNA sequences encoding GFP–cpScFv–cPIL and GFP–cpScFv were synthesized (Tsingke), PCR-amplified, and cloned downstream of the pUBQ10 promoter between the HindIII and NcoI restriction sites. For the MS2-based reporter, the 24× MS2 cassette was inserted into the 3′ UTR of Actin2, generating an Actin2–24× MS2 fusion construct driven by the UBQ10 promoter. In the same plasmid, the DNA fragment encoding mCherry–MCP was also cloned under the control of the UBQ10 promoter. For endogenous RNA imaging (including NtACT7, ACT2, FT, IRX9, ATPP2-A1, and CORTEX4), gRNA sequences were cloned into the pUBQ10::dCas13X–SunTag–IRES–GFP–cpScFv–cPIL construct, downstream of the AtU6-26 promoter. To visualize plasmodesmata, the CS of the specific PD marker PDCB1 (AT5G61130, Plasmodesmata Callose-Binding Protein 1), fused to mTagBFP2, was cloned under the control of the UBQ10 promoter. Unless otherwise stated, DNA polymerases, restriction endonucleases, ligases, and assembly mixes were purchased from New England Biolabs (NEB). All plasmids were verified by Sanger sequencing and introduced into Agrobacterium tumefaciens strain GV3101 for subsequent transient expression or stable transformation.

Agrobacterium-mediated infiltration of Nicotiana benthamiana

Agrobacterium tumefaciens strains were streaked onto LB agar plates supplemented with kanamycin (50 µg mL⁻¹) and rifampicin (25 µg mL⁻¹). Single colonies were inoculated into liquid LB medium containing the same antibiotics and grown overnight. Cells were harvested by centrifugation, resuspended in infiltration buffer (10 mM MES, pH 5.6; 10 mM MgCl₂; 150 µM acetosyringone), and adjusted to an OD₆₀₀ of approximately 0.6. After incubation for 2 h, the bacterial suspensions were infiltrated into the abaxial side of leaves of 3-week-old N. benthamiana plants using a needleless syringe. For co-infiltration, bacterial suspensions were mixed at a 1:1 (v/v) ratio. Plants were maintained in a growth chamber for 2 d after infiltration, and infiltrated leaf tissues were subsequently harvested for confocal microscopy.

Proteome-wide degron screening (LC–MS/MS)

Leaves of five-week-old tobacco (Nicotiana benthamiana) were treated with 10 µM MG132 or a mock solution (1× PBS) for 24 h. After treatment, leaf tissues were subjected to liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis (Shanghai Personal Biotechnology Co., Shanghai, China). Total protein was extracted using the FASP method, digested with trypsin (1:50 w/w), and analyzed on an Orbitrap Astral mass spectrometer (Thermo Fisher Scientific) coupled to a Vanquish Neo LC system. Data were acquired in Data-Independent Acquisition (DIA) mode and processed using DIA-NN software (v.1.8). Differentially expressed proteins were identified based on a fold-change >2 and FDR <0.01.

Assessment of degron-mediated degradation efficiency in Nicotiana benthamiana

To evaluate the degradation efficiency of candidate degron sequences, we cloned 10 previously reported protein degrons (S2 Table) as well as 90-nt fragments corresponding to 30 candidate C-terminal degron genes identified by proteomics into the C terminus of the GFP CDS in a 2A self-cleaving co-expression vector (S3 Table). This construct co-expresses a GFP-fusion reporter together with mCherry as an internal fluorescence control.

In addition, multiple truncated variants of CP13 were generated and inserted into the same reporter backbone (S4 Table). The resulting plasmids were introduced into Nicotiana benthamiana leaves by Agrobacterium-mediated infiltration. At 48 h post-infiltration, fluorescence signals were examined by fluorescence microscopy, and degron activity was quantified by comparing the GFP fluorescence intensity to the corresponding mCherry signal to normalize for expression variability.

Degron-dependent turnover kinetics of GFP fusion proteins

GFP–cpScFv or GFP–cpScFv–cPIL was transiently expressed in Nicotiana benthamiana leaves by Agrobacterium-mediated infiltration under the control of the same promoter. At 36 h post-infiltration, when robust fluorescence was readily detectable in the positive control, chemical treatments were initiated. To assess protein turnover, de novo protein synthesis was inhibited with 100 µM cycloheximide (CHX) prepared in 1× PBS. Infiltrated leaf tissues were harvested from the same infiltrated areas at 0, 30, 60, and 90 min after CHX treatment, and the collected samples were used for subsequent immunoblot analysis.

gRNA design and guide selection for dCas13X imaging

Guide RNAs were designed using a fixed Cas13X-compatible architecture consisting of the Cas13X direct repeat and a 30-nt spacer. For each target transcript, candidate target sites within the CDS were first identified in silico using Cas13design as a heuristic resource to prioritize regions with favorable predicted accessibility. For each target transcript, the five highest-scoring 23-nt candidate spacers predicted by Cas13design were selected for further processing. Each 23-nt candidate spacer was mapped back to the corresponding target transcript by reverse-complement alignment. The target site was then extended using authentic adjacent transcript sequence to generate a 30-nt Cas13X spacer. In this process, we prioritized designs in which the original 23-nt candidate core occupied positions 8–30 of the final 30-nt spacer, thereby retaining the Cas13design-predicted core within the region corresponding to the reported Cas13X crRNA 16–30 nt activity-sensitive region [30]. These candidates were then evaluated for predicted off-target risk by sequence-similarity analysis against the Arabidopsis TAIR10 transcript set. The candidate with the lowest predicted off-target risk was selected for gRNA construction and imaging validation. The total number of predicted target sites together with the guide scores, standardized guide scores, and predicted off-target information for the guides used in this work are provided in S5 Table.

RT-PCR

Total RNA was isolated using a Plant RNA Extraction Kit (BIOFIT, RN33050) and treated to remove residual genomic DNA according to the manufacturer’s instructions. Equal amounts of RNA were reverse-transcribed into cDNA using a reverse transcription kit (abm, G592). Quantitative RT-PCR (qRT-PCR) was then performed with gene-specific primers synthesized for the target genes. Each sample was analyzed with at least three technical replicates, and expression levels were normalized to a stably expressed internal reference gene. Relative transcript abundance was calculated using the 2^−ΔΔCt method. Primer sequences are provided in S6 Table.

Immunoblotting

Total protein was extracted from leaf tissues using extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% NP-40, 1 mM EDTA, and protease inhibitor cocktail). Lysates were clarified by centrifugation at 13,000g, separated by SDS–PAGE, and transferred to PVDF membranes. Membranes were probed with anti-GFP (1:5,000, Abbkine, ABT2020), anti-ACT2 (1:3,000, Agrisera), anti-GRP7 (1:2,000, PhytoAB, PHY3535A), anti-GAPDH (1:1,000, Agrisera), and anti-FLAG tag (1:1,000, Sigma) antibodies, followed by HRP-conjugated secondary antibodies. Signals were visualized using ECL substrate (Bio-Rad) and quantified using ImageJ. For stability assays, leaf discs were treated with 100 µM cycloheximide (CHX) or 50 µM MG132 and harvested at the indicated time points.

Single-molecule fluorescence in situ hybridization (smFISH)

smFISH was performed using Cy3-labeled Stellaris probes (Biosearch Technologies) targeting the ACT2 CDS, designed to exclude the dCas13X-binding region (S7 Table). A single construct co-expressing the dCas13X–SunTag/GFP–cpScFv–cPIL system and gRNA was transformed into Arabidopsis. Leaf tissues were fixed in 4% paraformaldehyde (PFA) in MES buffer (pH 5.7) for 30 min at room temperature. To facilitate permeabilization while maintaining tissue architecture, fixed samples were mounted on slides and processed via the freeze-cracking method using liquid nitrogen. After air-drying for 30 min, tissues were permeabilized in 70% ethanol with gentle agitation for ≥1 h and rehydrated in MES buffer. Hybridization was conducted overnight at 37 °C in a humidified dark chamber using 250 nM probes. Stringency washes were performed twice with 1× SSC containing 40% formamide at 37 °C for 45 min each to remove nonspecific binding. Samples were mounted in ProLong Diamond Antifade Mountant with DAPI (Invitrogen) for nuclear counterstaining prior to microscopy.

Generation of transgenic lines and micrografting

Stable transgenic Arabidopsis lines were generated via the Agrobacterium-mediated floral dip method using strain GV3101. T1 transformants were selected on 1/2 MS plates containing 50 µg mL⁻¹ kanamycin. Homozygous T3 lines with single-locus insertions were established for all experiments.

Micrografting was performed on 7-day-old seedlings grown vertically under short-day conditions (8 h light/16 h dark) to promote hypocotyl elongation. Scions were cut transversely and aligned with rootstocks using a silicone microtube (0.3 mm i.d.) for stabilization. Grafted plants were maintained in high humidity for 5–7 d to facilitate union formation. Adventitious roots were manually removed to ensure rootstock-dependent growth.

RNA pull-down–LC–MS/MS assay

FT sense mRNA and an antisense transcript (negative control) were synthesized by in vitro transcription using the MEGAscript T7 kit (Invitrogen) following the manufacturer’s instructions. Purified RNAs were 3′-end-labeled with desthiobiotin (RNA 3′ Desthiobiotinylation Kit; Thermo Scientific) for streptavidin-mediated capture. The concentration of each labeled RNA was quantified, and its integrity was verified by denaturing agarose gel electrophoresis to ensure comparable bait quality between reactions.

Total protein lysates were prepared in pre-chilled Pierce IP Lysis Buffer (Thermo Scientific), cleared by centrifugation at 4 °C, and normalized to equal input amounts. Where indicated, lysates were pre-cleared with magnetic beads to reduce nonspecific binding. RNA pull-down was performed using the Pierce Magnetic RNA–Protein Pull-Down Kit (Thermo Scientific). Desthiobiotinylated RNA was immobilized on streptavidin magnetic beads and incubated with protein lysate at 4 °C for 6 h with gentle rotation. Beads were washed stringently according to the kit protocol, and bound proteins were eluted. Eluates were divided for LC–MS/MS identification and immunoblot validation. All procedures were conducted on ice or at 4 °C to preserve RNA–protein interactions.

Microscope settings and image acquisition

Imaging was performed on an Olympus SpinSR10 spinning-disk confocal microscope equipped with a scanning unit and an EMCCD camera. A 63× oil-immersion objective (NA 1.2) was used for all quantification. GFP was excited at 488 nm (emission 500–550 nm), mCherry/Cy3 at 561 nm (emission 570–620 nm), and BFP at 405 nm (emission 440–450 nm). For Z-stacks, optical sections were acquired at 0.2 µm intervals. Laser power and exposure times were kept constant across all comparative experiments (typically 100–200 ms exposure).

Image processing and quantification of single RNA puncta

Image analysis was performed using the Fiji software [45]. The signal-to-noise ratio was calculated as “SNR = I_(spot)/I_(cyto)”, where “I_(spot)” is the mean intensity of an RNA punctum and “I_(cyto)” is the mean intensity of the neighboring cytoplasm.

To quantify the number of RNA molecules per cell, 3D Z-stacks were first converted to two-dimensional images using maximum-intensity projection (MIP). Regions of interest (ROIs) corresponding to individual cells were manually segmented based on the brightfield images. For the analysis of root marker genes in Fig 5, cellular ROIs were selected from the anatomically corresponding cell type for each marker on the basis of root tissue morphology in the brightfield images, namely cortical cells for CORTEX4, xylem cells for IRX9, and phloem-domain cells for ATPP2-A1. Background subtraction was applied using a rolling-ball algorithm with a radius of 50 pixels to remove inhomogeneous background illumination. Single RNA spots were detected and counted using the TrackMate plugin [46] within Fiji. The Laplacian of Gaussian (LoG) detector was employed, which is optimized for diffraction-limited spot detection. The estimated blob diameter was set to 0.5 μm, corresponding to the theoretical resolution limit of the optical system, and an intensity threshold was applied to exclude background noise. The number of RNA spots within each cellular ROI was automatically calculated and exported for analysis.

Quantitative reanalysis of the FT time-lapse imaging data

For dynamic analysis of FT mRNA trafficking, time-lapse image series were analyzed frame by frame in Fiji/ImageJ. A plasmodesmata-associated docking event was defined as the first frame in which an FT punctum became stably associated with the plasmodesmata-associated cell interface region. Dwell time was defined as the time interval between initial docking and the onset of intercellular transfer. A successful transfer event was defined as an FT punctum that moved from one cell side to the adjacent cell side across the same interface region and remained detectable after crossing. Transit time was defined as the interval between the onset of boundary crossing and completion of signal appearance on the adjacent cell side. Transfer frequency was calculated as the number of successful transfer events per minute per cell interface. In the grp7 background, puncta that reached plasmodesmata-associated regions but did not cross the cell boundary during the 3-min imaging period were scored as non-transferring docking events.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 9.0 (GraphPad Software). Data are presented as mean ± s.d. unless otherwise stated. Comparisons between two groups were performed using a two-tailed Student t test. Multi-group comparisons were analyzed using one-way ANOVA with Tukey’s post hoc test. For each dataset, the sample size (“n”) and the nature of the quantified units are specified in the corresponding figure legend. Unless otherwise stated, “n” refers to biologically independent samples, such as independent plants, roots, grafts, infiltrations, or cells derived from independent biological samples. P < 0.05 was considered statistically significant. Significance levels are indicated as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

S2 Fig. Screening of circularly permuted GFP–scFv mutants stabilized by dCas13X–SunTag.

(A) Representative fluorescence images of N. benthamiana epidermal cells expressing circularly permuted scFv mutants (mscFv) alone, or co-expressing each mscFv mutant with dCas13X–SunTag. Scale bars, 50 μm. (B) Quantification of the average number of fluorescent puncta per cell in tobacco epidermal cells co-expressing dCas13X–SunTag with GFP–cpScFv or wild-type GFP–scFv. Each dot represents one cell, and bars indicate mean ± s.d. (n = 30 cells from 3 biologically independent infiltrations). The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004043.s002

(TIF)

S4 Fig. Effect of the GFP–cpScFv–cPIL system on ACT2 RNA.

(A) RT–qPCR analysis of ACT2 mRNA abundance. Data are presented as mean ± s.d. (n = 3 biologically independent samples). (B) Immunoblot analysis of ACT2 protein levels (~42 kDa). GAPDH (~37 kDa) served as a loading control. (C) mRNA stability assay. ACT2 mRNA decay was measured following actinomycin D treatment (n = 3 biologically independent samples, mean ± s.d.). The data underlying this figure can be found in S1 Data. The uncropped blot and gel images are provided in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004043.s004

(TIF)

S6 Fig. RT-PCR and sequencing validate long-distance transport of FT mRNA, related to Fig 6. (A) RT-PCR detection of FT mRNA.

(B) Sanger sequencing confirmation of the FT amplicon obtained from ft_TG scions. The schematic indicates the position of the sequencing primer (7,037–7,060) and the covered region; representative chromatograms and sequence alignment show that the PCR product matches the FT transcript sequence, confirming that an intact and authentic FT mRNA is present in ft_TG scions. The uncropped blot and gel images are provided in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004043.s006

(TIF)

S7 Fig. Molecular identification of Arabidopsis T-DNA insertion mutants.

(A) Genotyping of T-DNA insertion mutants by three-primer PCR. Primers were designed using the T-DNA Primer Design platform. For each sample, the left lane corresponds to the PCR product amplified with LP + RP, whereas the right lane corresponds to the product amplified with LB + RP. (B) RT-PCR analysis of the corresponding target gene transcripts in the mutants. The uncropped blot and gel images are provided in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004043.s007

(TIF)