The dehydrin protein COR15 enhances antiviral RNA silencing by preventing viral protein-mediated SGS3 degradation and promoting SGS3 condensate formation

Abstract

RNA silencing, a conserved gene-regulatory mechanism mediated by small interfering RNAs (siRNAs), is a major component of plant antiviral immunity. The plant RNA-binding protein SUPPRESSOR OF GENE SILENCING 3 (SGS3) forms condensates that drive siRNA body assembly and promote siRNA biogenesis however, the regulators of SGS3 condensate formation and their roles in antiviral responses remain largely unknown. Here, we show that the cold-regulated 15-kDa dehydrin protein (COR15) homologs from Citrus aurantifolia and Nicotiana benthamiana interact with the citrus tristeza virus (CTV)-encoded p20 protein. COR15 abolishes the RNA silencing suppressor activity of p20, enhances antiviral RNA silencing, and confers resistance to CTV in N. benthamiana. COR15 also restricts infection by other positive-sense RNA viruses, suggesting that it functions as a broad-spectrum antiviral factor. Mechanistically, CTV infection induces NbCOR15 expression and promotes the formation of COR15 granules in the cytoplasm. COR15 attenuates p20-mediated SGS3 degradation by disrupting the p20–SGS3 interaction, thereby stabilizing SGS3. COR15 is recruited into siRNA bodies and undergoes liquid–liquid phase separation within these structures. Notably, COR15 promotes SGS3 condensate formation and efficient siRNA processing. CRISPR–Cas9-mediated knockout of COR15 in N. benthamiana plants reduces the number of SGS3 condensates, leads to abnormally sized SGS3 condensates, and impairs SGS3-dependent siRNA synthesis, indicating that COR15 is a previously unrecognized component of siRNA bodies. Together, our findings reveal the function and mechanism of a dehydrin protein in plant antiviral immunity and provide new insights into the molecular arms race underlying plant–virus co-evolution.

This study demonstrates that the plant dehydrin protein COR15 counteracts degradation of the RNA-silencing factor SGS3 mediated by the citrus tristeza virus-encoded p20 protein by disrupting the p20–SGS3 interaction. In addition, COR15 is recruited into SGS3 condensates and promotes small interfering RNA (siRNA) body assembly, thereby enhancing siRNA processing and antiviral immunity. These findings identify a previously unrecognized host factor and reveal a new mechanism underlying the plant–virus arms race.

Introduction

RNA silencing is a conserved small RNA (sRNA)-mediated gene-regulatory mechanism in eukaryotes (Lopez-Gomollon and Baulcombe, 2022). During viral invasion, host Dicer-like enzymes (DCLs) recognize double-stranded RNAs (dsRNAs) generated during viral replication or transcription and process them into virus-derived small interfering RNAs (vsiRNAs). These vsiRNAs are then methylated and loaded into Argonaute (AGO) proteins to assemble RNA-induced silencing complexes (RISCs). vsiRNA-guided RISCs mediate post-transcriptional gene silencing by cleaving complementary viral RNAs or repressing their translation (Lopez-Gomollon and Baulcombe, 2022; Li et al., 2025). In plants, RNA-dependent RNA polymerase 6 (RDR6) cooperates with its partner SUPPRESSOR OF GENE SILENCING 3 (SGS3) to amplify RNA silencing through the formation of siRNA bodies, which are critical for converting single-stranded RNA into double-stranded RNA and for the subsequent synthesis of secondary vsiRNAs (Mourrain et al., 2000; Peragine et al., 2004; Yoshikawa et al., 2005, 2013, 2021). Recent work has shown that SGS3 contains two prion-like domains and undergoes liquid–liquid phase separation (LLPS), which drives the formation of siRNA bodies (Kim et al., 2021; Han et al., 2023; Tan et al., 2023). However, the regulators of SGS3 condensate formation and their roles in sRNA biosynthesis and plant antiviral immunity remain elusive.

Plant viruses are widespread among cultivated crops and pose a serious threat to global food security (Scholthof et al., 2011; Guo et al., 2019). To successfully invade host plants, most viruses encode viral suppressors of gene silencing (VSRs), which counteract antiviral RNA silencing by disrupting key steps in the RNA silencing pathway (Guo et al., 2019; Lopez-Gomollon and Baulcombe, 2022). Citrus tristeza virus (CTV), a phloem-limited virus belonging to the genus Closterovirus within the family Closteroviridae, is one of the most economically significant viral pathogens affecting citrus production worldwide (Moreno et al., 2008; Fuchs et al., 2020; Sun et al., 2024). The 19.3-kb single-stranded, positive-sense RNA genome of CTV comprises 12 open reading frames (ORFs) (Moreno et al., 2008). ORF11 encodes the p20 protein, a VSR (Lu et al., 2004). Our previous study revealed that p20 interacts with both SGS3 and autophagy-related protein 8 (ATG8), forming ATG8–p20–SGS3 ternary complexes that sequester SGS3 into autophagosomes for degradation (Zhang et al., 2025). During the long-term arms race between plants and viruses, plants have evolved additional counter-defense strategies that target VSRs, thereby suppressing their activity or activating robust immunity. For example, plants use autophagy and the 26S proteasome pathway to degrade VSRs, such as rice stripe virus P3, cucumber mosaic virus 2b, tomato yellow leaf curl China virus βC1, and tobacco etch virus HC-Pro (Nakahara et al., 2012; Shen et al., 2016; Zhang et al., 2020; Tong et al., 2023). In addition, the helicase domain of tobacco mosaic virus p126 and the nonstructural proteins (NSs) of tomato spotted wilt virus are recognized by the nucleotide-binding leucine-rich repeat receptors N and Tsw, respectively, triggering hypersensitive responses (Erickson et al., 1999; Margaria et al., 2007). However, whether and how plants deploy counter-counter-defense strategies against CTV p20-mediated perturbation of RNA silencing remain unknown.

Dehydrins, which belong to the group 2 late embryogenesis abundant (LEA) protein family, are induced by diverse environmental factors, including drought, cold, salinity, and heavy metals (Graether and Boddington, 2014). Dehydrins also respond to pathogen infection (Ventelon-Debout et al., 2004; Tan et al., 2013; Musser et al., 2014; Rosales et al., 2014; Tian et al., 2019); however, their roles in plant resistance to biotic stress remain poorly understood. Heterologous expression of wheat (Triticum aestivum L.) dehydrin DHN-5 enhances the resistance of Arabidopsis thaliana to Botrytis cinerea and Alternaria solani (Drira et al., 2015, 2016). In contrast, overexpression of Arachis duranensis dehydrin AdDHN1 in A. thaliana increases plant susceptibility to Meloidogyne incognita (Mota et al., 2019). Nevertheless, whether dehydrins participate in plant antiviral immunity, and the mechanisms by which they may do so, remain unknown.

Here, we report a previously unrecognized role of the plant dehydrin cold-regulated 15-kDa protein (COR15) in antiviral immunity against CTV. Knockout of COR15 increased CTV accumulation, whereas COR15 overexpression reduced CTV accumulation in Nicotiana benthamiana plants. COR15 interacted with p20 and SGS3 and disrupted the p20–SGS3 interaction, thereby protecting SGS3 from autophagic degradation. We further demonstrated that COR15 was recruited into phase-separated SGS3 condensates and contributed to their formation. This study reveals a dehydrin-mediated plant counter-counter-defense mechanism during plant–virus co-evolution and identifies a hitherto unknown factor that modulates SGS3 condensate formation.

Results

COR15 homologs from Citrus aurantifolia and Nicotiana benthamiana interact with CTV p20

To identify host proteins that interact with p20, we screened a cDNA library constructed from mRNA isolated from CTV-infected C. aurantifolia leaves using a yeast two-hybrid (Y2H) assay with p20 as bait. A cDNA encoding dehydrin COR15 was identified. We then cloned the full-length coding sequences of CaCOR15 and NbCOR15 from C. aurantifolia and N. benthamiana, respectively. Split-ubiquitin-based membrane yeast two-hybrid (MYTH) assays showed that both CaCOR15 and NbCOR15 interacted with p20 in yeast cells (Figure 1A). To confirm the interaction between p20 and COR15 in vivo, we performed co-immunoprecipitation (co-IP) assays and found that p20 co-immunoprecipitated with CaCOR15-YFP and NbCOR15-YFP but not with YFP alone (Figure 1B). In bimolecular fluorescence complementation (BiFC) and luciferase complementation imaging (LCI) assays, co-expression of p20 with either CaCOR15 or NbCOR15 in N. benthamiana leaves produced interaction signals, whereas no signal was detected in the negative controls (Figures 1C and 1D). Subcellular localization assays showed that CaCOR15 localized to both the nucleus and cytoplasm (Supplemental Figure 1). Interestingly, compared with the empty vector (EV) and mock-infiltrated controls, p20 expression and CTV infection induced CaCOR15 to form granule-like bodies in the cytoplasm, and this effect was not attributable to differences in the abundance of CaCOR15-CFP protein (Figure 1E–1J). Furthermore, RT–qPCR analysis showed that NbCOR15 mRNA levels in CTV-infected plants were approximately 4.3- and 11.0-fold higher than those in mock-treated plants at 5 and 14 days post-inoculation (dpi), respectively (Supplemental Figure 2). Together, these results demonstrate that COR15 interacts with p20 and may be involved in the host response to CTV infection.

Figure 1.

CTV p20 interacts with CaCOR15 and NbCOR15.

(A) Split-ubiquitin yeast two-hybrid assay showing the interactions of p20 with CaCOR15 and NbCOR15. Yeast cultures co-transformed with LargeT-Cub and NubG-p53 and those co-transformed with LargeT-Cub and Ost1-NubG were used as positive and negative controls, respectively.

(B) Co-immunoprecipitation (co-IP) assay detecting the interaction of CTV p20 with CaCOR15 and NbCOR15 in N. benthamiana leaves. The red asterisk indicates the p20–CFP–HA band.

(C) Bimolecular fluorescence complementation (BiFC) assay showing the interactions of CTV p20 with CaCOR15 and NbCOR15. Scale bars, 20 μm. The merged panel shows the merged YFP fluorescence and bright-field images.

(D) Luciferase complementation imaging (LCI) assay showing the interactions of CTV p20 with CaCOR15 and NbCOR15. cps, counts per second.

(E) Subcellular localization of CaCOR15-CFP in N. benthamiana leaves infiltrated with EV or p20. Scale bars, 20 μm.

(F) Relative density of CaCOR15-CFP granules in N. benthamiana leaves infiltrated with EV or p20.

(G) Immunoblotting analysis of protein expression in the samples shown in (E).

(H) Subcellular localization of CaCOR15-CFP in mock-inoculated and CTV-infected N. benthamiana leaves. Scale bars, 20 μm.

(I) Relative density of CaCOR15-CFP granules in mock-inoculated and CTV-infected N. benthamiana leaves.

(J) Immunoblotting analysis of protein expression in the samples shown in (H).

Values in (F) and (I) represent means ± SD from three independent experiments. Significant differences were determined using a one-tailed Student’s t-test (∗∗p < 0.01, ∗∗∗p < 0.001). In (G) and (J), Coomassie Brilliant Blue staining of the Rubisco large subunit was used as a loading control, and relative band intensities were calculated using ImageJ software.

COR15 functions as an antiviral factor during CTV infection

To investigate the role of COR15 during CTV infection, we transiently overexpressed NbCOR15-FLAG in leaves of CTV-infected N. benthamiana plants and measured the levels of viral genomic RNA (gRNA) and coat protein (CP). Compared with the EV control, NbCOR15-FLAG overexpression reduced viral gRNA and CP levels in N. benthamiana plants (Supplemental Figure 3). To further assess the function of NbCOR15 during CTV infection, we systemically silenced NbCOR15 using tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS). Relative NbCOR15 mRNA levels in TRV:NbCOR15-infected plants decreased to approximately 43% of those in TRV:00-infected plants (Supplemental Figure 4A). When the CTV infectious clone was inoculated into TRV:00- and TRV:NbCOR15-infected N. benthamiana plants, more severe leaf crinkling symptoms developed in newly emerged leaves of NbCOR15-silenced plants than in control plants at 21 dpi (Supplemental Figure 4B). Consistently, viral gRNA and CP levels were significantly higher in NbCOR15-silenced plants than in control plants (Supplemental Figures 4C and 4D).

We then generated NbCOR15- knockout N. benthamiana mutants (cor15) using CRISPR-Cas9-mediated genome editing. Sequencing revealed a 1-bp deletion and a 1-bp insertion at the gRNA target site in the cor15-4 and cor15-6 lines, respectively (Supplemental Figure 5A). The mutant plants showed no obvious phenotypic differences from control plants under normal growth conditions (Supplemental Figure 5B). Following CTV infection, cor15-4 and cor15-6 plants exhibited more severe leaf crinkling and curling symptoms than wild-type (WT) plants at 21 dpi (Figure 2A). qPCR and northern blot analyses revealed higher CTV gRNA levels but lower vsiRNA accumulation in cor15-4 and cor15-6 plants than in WT plants (Figures 2B and 2C). We next calculated relative vsiRNA/gRNA ratios to evaluate vsiRNA amplification efficiency in these plants. The CTV-vsiRNA/gRNA ratios were markedly lower in cor15 mutants than in WT plants (Figure 2D). Moreover, immunoblotting showed that CTV CP accumulated to significantly higher levels in the cor15-4 and cor15-6 lines than in WT plants (Figure 2E). Collectively, these results indicate that cor15 mutant plants are impaired in vsiRNA amplification and CTV resistance.

Figure 2.

COR15 negatively regulates CTV infection.

(A) Symptoms of wild-type (WT) and NbCOR15-knockout (cor15-4 and cor15-6) plants inoculated with CTV at 21 days post-inoculation (dpi).

(B) Relative accumulation of CTV genomic RNA (gRNA) in CTV-infected WT and NbCOR15-knockout plants at 21 dpi.

(C) Northern blot analysis of the accumulation of CTV vsiRNAs in systemic leaves of CTV-infected WT and NbCOR15-knockout plants at 21 dpi.

(D) Relative CTV vsiRNA/gRNA ratios calculated from the data shown in (B) and (C).

(E) Relative accumulation of CTV CP in CTV-infected WT and NbCOR15-knockout plants at 21 dpi.

(F) Symptoms of WT and CaCOR15-expressing (CaCOR15-3 and CaCOR15-6) plants inoculated with CTV at 30 dpi.

(G) Relative accumulation of CTV gRNA in CTV-infected WT and CaCOR15-expressing plants at 30 dpi.

(H) Northern blot analysis of the accumulation of CTV vsiRNAs in systemic leaves of CTV-infected WT and CaCOR15-expressing plants at 30 dpi.

(I) Relative CTV vsiRNA/gRNA ratios calculated from the data shown in (G) and (H).

(J) Relative accumulation of CTV CP in CTV-infected WT and CaCOR15-expressing plants at 30 dpi.

The NbActin gene was used as an internal control in RT–qPCR assays. For immunoblotting, CTV CP was detected using an anti-CTV CP antibody. Ethidium bromide-stained ribosomal RNA (rRNA) and Coomassie Brilliant Blue-stained Rubisco large subunit served as loading controls for RNA (mRNA and siRNA) and protein analyses, respectively. Band intensities were quantified using ImageJ software. Values represent means ± SD from three independent experiments. Significant differences were determined using a one-tailed Student’s t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001).

To investigate the role of CaCOR15 in CTV defense, we transiently overexpressed CaCOR15-FLAG in CTV-infected C. sinensis leaves. Infiltrated leaves were collected at 5 dpi and analyzed by RT–qPCR and immunoblotting. The results showed that transient overexpression of CaCOR15 significantly reduced CTV accumulation in CTV-infected C. sinensis leaves (Supplemental Figures 6A–6C). We then silenced CsCOR15 in CTV-infected C. sinensis leaves using a TRV vector. qPCR analysis showed that CsCOR15 expression was significantly downregulated at 7 dpi, and CsCOR15 silencing significantly increased CTV accumulation in citrus leaves (Supplemental Figures 6D–6F). We further generated CaCOR15-expressing transgenic lines (CaCOR15-3 and CaCOR15-6) in WT N. benthamiana. CaCOR15 expression was verified in T2-generation transgenic plants by RT–PCR (Supplemental Figure 7). Upon CTV inoculation, CaCOR15-expressing plants showed milder symptoms than WT plants (Figure 2F) and accumulated lower levels of CTV gRNA and CP but higher levels of CTV-derived vsiRNAs (Figures 2G, 2H, and 2J). The CTV-vsiRNA/gRNA ratios increased by at least 4.7-fold in CaCOR15-expressing plants compared with WT plants (Figure 2I). Together, these results demonstrate that COR15 positively regulates vsiRNA amplification and plant resistance to CTV.

COR15 promotes antiviral RNA silencing

Given that p20 suppresses RNA silencing and COR15 interacts with p20, we hypothesized that COR15 confers CTV resistance by participating in the antiviral RNA-silencing pathway. We co-expressed GFP with p20-FLAG and CaCOR15-FLAG, p20-FLAG and EV, CaCOR15-FLAG alone, or EV alone in leaves of N. benthamiana 16c plants. At 2 dpi, GFP signals were observed in all infiltrated patches, although relatively weaker GFP signals were detected in patches co-expressing CaCOR15-FLAG and GFP (Figure 3A). At 5 dpi, GFP signals were almost undetectable in CaCOR15-FLAG- and EV-infiltrated patches, whereas leaf patches co-infiltrated with p20-FLAG and EV or with p20-FLAG and CaCOR15-FLAG still showed visible GFP signals. Notably, GFP fluorescence was much weaker in patches co-infiltrated with p20-FLAG and CaCOR15-FLAG than in those co-infiltrated with p20-FLAG and EV (Figure 3A). Immunoblot and northern blot analyses showed that GFP protein and mRNA levels were markedly reduced, whereas GFP-derived siRNA levels were significantly increased, in patches co-expressing p20-FLAG and CaCOR15-FLAG compared with those co-expressing p20-FLAG and EV (Figure 3B–3D). These results suggest that CaCOR15 promotes local RNA silencing and inhibits the VSR activity of p20. Phylogenetic and sequence analyses further revealed that COR15 homologs are widely distributed across monocot and dicot plants and share high sequence similarity in the conserved K and S segments (Supplemental Figures 8A and 8B). To investigate whether COR15 homologs also contribute to RNA silencing, we co-infiltrated hairpin GFP (hpGFP) and GFP with CaCOR15, NbCOR15, or their homologs from Triticum aestivum (TaHIRD11), Zea mays (ZmHIRD11), Glycine max (GmHIRD11), and Solanum lycopersicum (SlHIRD11) into WT N. benthamiana leaves. All tested COR15 homologs enhanced GFP silencing, mirroring the effect of CaCOR15 (Supplemental Figures 8C–8E).

Figure 3.

COR15 promotes RNA silencing.

(A) Effect of CaCOR15 on the VSR activity of p20. N. benthamiana 16c leaves co-infiltrated with different vector combinations were visualized under UV light at 2 and 5 dpi. The images are representative of at least 50 individual leaf samples from three independent experiments.

(B–D) GFP mRNA (B), GFP-derived siRNA (C), and protein (D) levels in the infiltrated patches from (A) at 5 dpi.

(E) Effect of NbCOR15 knockout on local GFP silencing. The GFP construct was infiltrated into the leaves of WT and NbCOR15-knockout (cor15-4 and cor15-6) N. benthamiana plants. Photographs were taken under UV light at 3 dpi.

(F–H) GFP mRNA (F), GFP-derived siRNA (G), and protein (H) levels in the infiltrated patches shown in (E).

(I) Effect of NbCOR15 knockdown on systemic GFP silencing. TRV:00- and TRV:NbCOR15-treated 16c plants were infiltrated with GFP, and GFP expression in the upper systemic leaves was observed under UV light at 15 dpi.

(J–L) GFP mRNA (J), GFP-derived siRNA (K), and protein (L) levels in the upper systemic leaves of the infiltrated plants shown in (I).

Ethidium bromide-stained ribosomal RNA (rRNA) and Coomassie Brilliant Blue (CBB)-stained Rubisco large subunit served as loading controls for RNA (mRNA and siRNA) and protein analyses, respectively. Band intensities were quantified using ImageJ software. Values represent means ± SD from three independent experiments. Significant differences were determined using a one-tailed Student’s t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant).

To further investigate whether COR15 participates in the RNA silencing pathway, we infiltrated GFP into the leaves of WT, cor15-4, and cor15-6 N. benthamiana plants. At 3 dpi, GFP signals were stronger in cor15-4 and cor15-6 leaves than in WT leaves, consistent with the observed changes in GFP mRNA, GFP-derived siRNA, and protein levels (Figure 3E–3H). Furthermore, we silenced NbCOR15 in N. benthamiana 16c plants using TRV-based VIGS and infiltrated GFP into NbCOR15-silenced and control 16c plants. At 15 dpi, leaf veins in the upper leaves of control plants turned red, whereas GFP signals remained strong throughout the upper leaves of NbCOR15-silenced 16c plants (Figure 3I). Consistently, NbCOR15 knockdown increased GFP mRNA and protein levels and reduced the biogenesis of GFP-derived siRNAs (Figure 3J–3L). These data indicate that NbCOR15 deficiency impedes RNA silencing.

COR15 protects SGS3 from p20-mediated degradation

Our previous study showed that p20 suppresses RNA silencing by mediating the autophagic degradation of SGS3 (Zhang et al., 2025). Here, we investigated whether COR15 affects SGS3 protein stability. NbCOR15-YFP-FLAG and YFP-FLAG were individually co-expressed with NbSGS3-CFP-HA in N. benthamiana leaves. Immunoblot and RT–qPCR analyses showed that NbCOR15 expression had no significant effect on NbSGS3 mRNA or protein levels (Supplemental Figure 9). We then examined the effect of COR15 on SGS3 accumulation in the presence of p20. Agrobacterium cultures carrying NbCOR15-Myc at OD₆₀₀ values of 0.1, 0.2, 0.4, and 0.8 were separately co-infiltrated with Agrobacterium cultures carrying p20-YFP-FLAG and NbSGS3-CFP-HA. At 3 dpi, NbSGS3-CFP-HA protein levels increased with increasing NbCOR15-Myc expression (Figure 4A), whereas the transcript levels of NbSGS3-CFP-HA did not differ significantly among treatments (Supplemental Figure 10A). These results indicate that NbCOR15 prevents p20-mediated NbSGS3 degradation. Furthermore, we co-infiltrated YFP-NbSGS3 and p20-FLAG or the EV into WT, cor15-4, and cor15-6 N. benthamiana leaves using a half-leaf inoculation method. The results showed that YFP-NbSGS3 protein levels were markedly lower in p20-infiltrated leaf patches than in EV-infiltrated WT, cor15-4, and cor15-6 leaf patches (Figure 4B). Notably, YFP-NbSGS3 degradation was more pronounced in cor15-4 and cor15-6 leaf patches than in WT leaf patches in the presence of p20 (Figure 4B and Supplemental Figure 10B). Consistently, protein stability assays showed that YFP-NbSGS3 was degraded more slowly in extracts expressing NbCOR15-Myc than in EV control extracts (Figure 4C). Notably, the protective effect of NbCOR15 on NbSGS3 was abolished by treatment with the autophagy inhibitor 3-methyladenine (3-MA) but not by the proteasome inhibitor MG132 (Figure 4C), further indicating that NbCOR15 prevents the p20-mediated autophagic degradation of NbSGS3.

Figure 4.

NbCOR15 interferes with the p20–SGS3 interaction to counteract p20-mediated SGS3 degradation.

(A) Effect of NbCOR15 on NbSGS3 protein accumulation in the presence of p20.

(B) Effect of NbCOR15 knockout on p20-mediated NbSGS3 degradation.

(C) Protein stability assays showing that NbCOR15 inhibits p20-mediated autophagic degradation of NbSGS3.

(D) Co-localization analysis of YFP-NbSGS3 with mCherry-NbATG8f-labeled autophagic bodies. YFP-NbSGS3, mCherry-NbATG8f, and p20-FLAG were co-infiltrated into N. benthamiana leaves together with NbCOR15-HA or EV. The infiltrated leaves were treated with E-64d at 48 hpi and examined 12 h after E-64d treatment. Scale bars, 20 μm. Co-localization was further assessed by analyzing the overlap of fluorescence signals in the regions enclosed by white dashed lines in the enlarged images.

(E) Statistical analysis of the percentage of YFP-NbSGS3 signals co-localized with mCherry-NbATG8f-labeled autophagic bodies in (D).

(F) Split-ubiquitin-based membrane yeast two-hybrid assay showing the interaction between NbCOR15 and NbSGS3. Yeast cultures co-transformed with LargeT-Cub and NubG-p53 and those co-transformed with LargeT-Cub and Ost1-NubG were used as positive and negative controls, respectively.

(G) LCI assay showing the interaction between NbCOR15 and NbSGS3. cps, counts per second.

(H) BiFC assay showing the interaction between NbCOR15 and NbSGS3. Scale bars, 20 μm.

(I) Co-IP assay showing the interaction between NbCOR15 and NbSGS3.

(J) Competitive co-IP assay showing the effect of NbCOR15 on the interaction between p20 and NbSGS3. p20-FLAG and YFP-NbSGS3 were co-expressed with NbCOR15-HA or CFP-HA as a control in N. benthamiana leaves. Total protein was extracted and immunoprecipitated with anti-FLAG beads, followed by immunoblotting using anti-FLAG, anti-GFP, and anti-HA antibodies.

(K) Competitive LCI assay showing the effect of NbCOR15 on the interaction between p20 and NbSGS3. p20-nLUC and cLUC-NbSGS3 were co-infiltrated into N. benthamiana leaves together with NbCOR15-FLAG or EV. cps, counts per second.

Coomassie Brilliant Blue (CBB) staining of the Rubisco large subunit was used as a loading control in (A)–(C). Band intensities were quantified using ImageJ software. Values in (A)–(C), (E), (J), and (K) represent means ± SD from three independent experiments. Significant differences were determined using one-way ANOVA with Tukey’s multiple-comparison test (p < 0.05) for (A), (J), and (K), and a one-tailed Student’s t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant) for (C) and (E).

Considering that p20 induces autophagy and hijacks SGS3 into autophagosomes for degradation (Zhang et al., 2025), we investigated whether COR15 influences p20-induced autophagy. Monodansylcadaverine (MDC) staining revealed no significant difference in autophagic activity between NbCOR15-FLAG- and EV-infiltrated leaves after treatment with DMSO or the chemical autophagy inducer benzo-(1,2,3)-thiadiazole-7-carbothioic acid (BTH), indicating that NbCOR15 does not regulate autophagic activity (Supplemental Figures 11A–11C). We further investigated whether NbCOR15 inhibits p20-induced autophagy. MDC staining showed that numerous autophagic structures were present in p20-expressing leaves but not in the EV control, consistent with the role of p20 as an autophagy activator (Zhang et al., 2025). Moreover, autophagic activity did not differ significantly between leaves co-expressing p20-FLAG and NbCOR15-YFP and those co-expressing p20-FLAG and YFP, suggesting that NbCOR15 does not inhibit p20-induced autophagy (Supplemental Figures 11D–11F). Furthermore, colocalization assays showed that transient expression of NbCOR15-FLAG significantly reduced the proportion of YFP-NbSGS3 signal that colocalized with mCherry-NbATG8f-labeled autophagic structures from 59% to 20% (Figures 4D and 4E). These results suggest that NbCOR15 prevents NbSGS3 from localizing to autophagosomes in the presence of p20.

We then verified the interaction between NbCOR15 and NbSGS3. MYTH assays showed that yeast cells co-transformed with NubG-NbCOR15 and NbSGS3-Cub grew well on SD/−Trp−Leu−His−Ade medium (Figure 4F). LCI and BiFC assays showed visible interaction signals in leaves co-infiltrated with NbCOR15 and NbSGS3 (Figures 4G and 4H). In co-IP assays, YFP-NbSGS3 was co-immunoprecipitated with NbCOR15-FLAG (Figure 4I). We further examined whether COR15 affects the interaction between p20 and SGS3. A competitive co-IP assay was performed by co-expressing p20-FLAG and YFP-NbSGS3 with NbCOR15-HA or CFP-HA in N. benthamiana leaves. Immunoblot analysis showed that both YFP-NbSGS3 and NbCOR15-HA were co-immunoprecipitated by p20-FLAG. Conversely, NbCOR15-HA, but not CFP-HA, reduced the level of YFP-NbSGS3 immunoprecipitated by p20-FLAG (Figure 4J). LCI assays also showed that NbCOR15-FLAG reduced the luminescence signals generated by the interaction between p20–N-terminal luciferase (nLUC) and C-terminal luciferase (cLUC)–NbSGS3 (Figure 4K). To determine how NbCOR15 interferes with the p20–NbSGS3 interaction, we generated four truncated NbSGS3 mutants by deleting the N-terminal region (NbSGS3dN), zinc finger domain (NbSGS3dZF), rice gene X and SGS3 domain (NbSGS3dXS), or coiled-coil domain (NbSGS3dCC) (Supplemental Figure 12A). BiFC assays were then performed to identify the regions mediating the interactions of NbSGS3 with p20 and NbCOR15. Deletion of either the N-terminal region or the CC domain markedly disrupted the NbSGS3–p20 interaction (Supplemental Figure 12B). Interestingly, the N-terminal region was also required for the NbSGS3–NbCOR15 interaction, whereas deletion of the CC domain did not alter fluorescence intensity but changed the subcellular localization of the NbSGS3–NbCOR15 complexes, impairing their ability to form condensates (Supplemental Figure 12C). These findings suggest that NbCOR15 may interfere with the p20–NbSGS3 interaction by competitively binding to the N-terminal region of NbSGS3. Collectively, these results demonstrate that COR15 disrupts the p20–SGS3 interaction, thereby preventing p20-mediated degradation of NbSGS3.

COR15 facilitates the formation of SGS3 condensates

NbCOR15-YFP localized to the cytoplasm and nucleus (Supplemental Figure 1), whereas NbSGS3-CFP localized to discrete granule-like structures in the cytoplasm (Supplemental Figure 13). Intriguingly, colocalized cytoplasmic aggregates were observed in N. benthamiana cells co-expressing NbCOR15-YFP and NbSGS3-CFP (Figure 5A). In plants, SGS3 and RDR6 form cytoplasmic SGS3/RDR6 bodies for siRNA amplification (Kumakura et al., 2009). Confocal microscopy showed that NbCOR15-CFP colocalized with SGS3/RDR6 bodies (Figure 5A), suggesting that the contribution of NbCOR15 to antiviral RNA silencing is associated with SGS3/RDR6 bodies. SGS3 undergoes phase separation to drive siRNA body assembly, which is crucial for endogenous gene silencing and antiviral immunity (Han et al., 2023; Tan et al., 2023). Time-lapse microscopy showed that aggregates containing both NbCOR15 and NbSGS3 were mobile and underwent fusion (Figure 5B), indicating that they exhibit phase-separation properties in vivo. In addition, fluorescence recovery after photobleaching (FRAP) assays showed that NbCOR15–NbSGS3 condensates rapidly recovered after photobleaching (Figures 5C and 5D), and treatment with 1,6-hexanediol, a chemical inhibitor that disrupts liquid-like droplets, efficiently dispersed NbCOR15–NbSGS3 interaction granules (Figures 5E and 5F). These results indicate that NbCOR15 is recruited into phase-separated SGS3 condensates.

Figure 5.

NbCOR15 facilitates the formation of SGS3 condensates.

(A) Co-localization analysis of NbCOR15 with NbSGS3 and SGS3/RDR6 bodies.

(B) Movement and fusion of NbCOR15–NbSGS3 co-localized condensates in N. benthamiana leaf epidermal cells. White arrows indicate condensates undergoing fusion. Scale bars, 10 μm.

(C) Representative fluorescence recovery after photobleaching images of NbCOR15–NbSGS3 interaction granules in N. benthamiana leaf epidermal cells.

(D) Recovery curve of photobleached NbCOR15–NbSGS3 granules in the cytoplasm. The fluorescence intensity of each granule was normalized to its pre-bleaching intensity. Data are presented as means ± SD from three independent experiments.

(E) Effect of 1,6-hexanediol on NbCOR15–NbSGS3 interaction granules in N. benthamiana leaf epidermal cells. Leaves were treated with H₂O or 15% 1,6-hexanediol for 5–15 min before imaging.

(F) Density of NbCOR15–NbSGS3 condensates in (E).

(G) Effect of NbCOR15 overexpression on the formation of SGS3 condensates.

(H) Density of SGS3 condensates in (G).

(I) Effect of NbCOR15 knockout on the formation of SGS3 condensates. NbSGS3-CFP was infiltrated into the leaves of WT and NbCOR15-knockout N. benthamiana plants and imaged at 2 dpi.

(J) Density of SGS3 condensates in (I).

(K) Size distribution of SGS3 condensates in (I).

(L) Effect of NbCOR15 transient overexpression on tasiR-ARF3 accumulation in mock-infiltrated and CTV-infected plants.

(M) Effect of NbCOR15 knockout on tasiR-ARF3 accumulation in mock-infiltrated and CTV-infected plants. For northern blot assays, ethidium bromide-stained ribosomal RNA (rRNA) served as a loading control for siRNAs, and relative band intensities were quantified using ImageJ software. Values in (D), (F), (H), and (K) represent means ± SD from three independent experiments. Significant differences were determined using a one-tailed Student’s t-test (∗∗p < 0.01, ∗∗∗p < 0.001). Scale bars in (A), (C), (E), (G), and (I) represent 20 μm.

We hypothesized that COR15 may be involved in the formation of SGS3 condensates. To test this hypothesis, we co-expressed NbSGS3-CFP driven by either its native promoter (proSGS3:NbSGS3-CFP) or the constitutive 35S promoter (35S:NbSGS3-CFP), with NbCOR15-YFP or YFP in N. benthamiana leaves. Confocal imaging showed that transient expression of NbCOR15 significantly increased the density of NbSGS3-CFP condensates (Figures 5G and 5H; Supplemental Figure 14), indicating that NbCOR15 promotes the formation of SGS3 condensates. We then separately infiltrated proSGS3:NbSGS3-CFP and 35S:NbSGS3-CFP into WT and NbCOR15-knockout N. benthamiana leaves. The density of SGS3 condensates was significantly lower in NbCOR15-knockout plants than in WT plants (Figures 5I and 5J; Supplemental Figure 15), and this decrease was not attributable to differences in NbSGS3-CFP protein levels (Supplemental Figure 16). Strikingly, the percentage of SGS3 condensates < 1 μm in diameter was markedly lower in NbCOR15-knockout plants, whereas NbCOR15 knockout resulted in a higher proportion of SGS3 condensates larger than 4 μm in diameter compared with WT plants (Figure 5K), reminiscent of the inactive aggregates driven by SGS3dXS, a rice gene X and SGS3 domain truncation mutant(Tan et al., 2023). These findings suggest that NbCOR15 contributes to the formation of SGS3 condensates.

We further examined the effect of NbCOR15 on siRNA body assembly and endogenous trans-acting siRNA (tasiRNA) biogenesis. BiFC and co-IP assays showed that transient expression of NbCOR15 enhanced the interaction between NbSGS3 and NbRDR6 (Supplemental Figure 17). Consistently, northern blot analysis revealed that the accumulation of tasiR-ARF3, an SGS3-dependent tasiRNA, was significantly higher in NbCOR15-expressing leaf patches than in EV-infiltrated patches (Figure 5L). RT–qPCR analysis revealed that transient expression of NbCOR15 significantly increased the accumulation of tasiR-ARF3 and reduced the transcript levels of its target genes NbARF2, NbARF3, and NbARF4 (Supplemental Figure 18). Consistently, knockout of NbCOR15 impaired tasiR-ARF3 biogenesis and increased the transcript levels of NbARF2, NbARF3, and NbARF4, especially during CTV infection (Figure 5M and Supplemental Figure 19). Together, these results indicate that COR15 is incorporated into SGS3 condensates and facilitates siRNA body assembly and SGS3-mediated siRNA biogenesis.

COR15-mediated antiviral activity depends on SGS3

To further elucidate the genetic relationship between COR15 and SGS3, we silenced NbSGS3 in wild-type N. benthamiana plants using TRV-based VIGS (Supplemental Figure 20) and co-infiltrated the CTV infectious clone together with NbCOR15-FLAG or EV into non-silenced and NbSGS3-silenced N. benthamiana leaves. Compared with non-silenced leaves, NbSGS3-silenced leaves accumulated higher levels of CTV gRNA and CP, suggesting that NbSGS3 functions as an antiviral factor during CTV infection (Supplemental Figure 21). Moreover, NbCOR15 overexpression inhibited CTV accumulation in non-silenced leaves, whereas this antiviral effect was compromised in NbSGS3-silenced leaves (Supplemental Figure 21). Furthermore, we silenced NbSGS3 in CaCOR15-expressing transgenic plants (Supplemental Figure 20) and performed CTV infection via agroinfiltration. We then compared the symptoms of non-silenced and NbSGS3-silenced WT and CaCOR15-expressing transgenic N. benthamiana plants upon CTV infection. Non-silenced CaCOR15-expressing plants showed attenuated symptoms compared with WT control plants (Figure 6A–6C). In contrast, CaCOR15-expressing plants showed similar degrees of CTV infection to WT plants in NbSGS3-silenced treatments (Figure 6A–6C). We then silenced NbSGS3 in NbCOR15-knockout N. benthamiana plants to generate SGS3-KD/cor15 plants (Supplemental Figure 20). Upon CTV infection, SGS3-KD/cor15 plants exhibited resistance similar to that of SGS3-KD plants (Figure 6D–6F), suggesting that NbSGS3 silencing attenuated NbCOR15-knockout-mediated susceptibility to CTV infection. Together, these results indicate that COR15 exerts its antiviral activity primarily through the SGS3-mediated RNA silencing pathway.

Figure 6.

SGS3 is required for COR15-mediated resistance to CTV infection.

(A) Symptoms of NbSGS3-silenced and non-silenced WT and CaCOR15-expressing N. benthamiana plants inoculated with CTV at 21 dpi.

(B and C) Relative accumulation of viral genomic RNA (gRNA) (B) and CP (C) in the CTV-infected plants shown in (A).

(D) Symptoms of WT, NbCOR15-knockout, NbSGS3-silenced, and NbCOR15/NbSGS3-deficient mutant plants inoculated with CTV at 21 dpi.

(E and F) Relative accumulation of viral gRNA (E) and CP (F) in the CTV-infected plants shown in (D).

The NbActin gene was used as an internal control in the RT-qPCR assays. For immunoblotting, Coomassie Brilliant Blue (CBB) staining of the Rubisco large subunit was used as a loading control, and band intensities were quantified using ImageJ software. Values represent means ± SD from three independent experiments. Significant differences were determined using a one-tailed Student’s t-test (∗p < 0.05, ∗∗p < 0.01; ns, not significant).

Overexpression of CaCOR15 reduces PVY and ToBRFV accumulation in N. benthamiana

To investigate whether COR15 exerts antiviral activity against other plant viruses, we agroinoculated WT and CaCOR15-expressing transgenic N. benthamiana plants with GFP-tagged potato virus Y (PVY; genus Potyvirus) or tomato brown rugose fruit virus (ToBRFV; genus Tobamovirus). Symptom observation revealed that CaCOR15 overexpression delayed the onset of GFP fluorescence following PVY-GFP and ToBRFV-GFP infection (Figures 7A and 7B). Consistently, RT–qPCR and immunoblot analyses showed that viral accumulation was significantly reduced in CaCOR15-expressing plants compared with WT plants (Figure 7C–7F). These results indicate that COR15 confers broad-spectrum antiviral activity against plant viruses.

Figure 7.

broad-spectrum antiviral activity in N. benthamiana.

(A) Symptoms of WT and CaCOR15-expressing N. benthamiana plants inoculated with PVY-GFP at 5 dpi.

(B) Symptoms of WT and CaCOR15-expressing N. benthamiana plants inoculated with ToBRFV-GFP at 7 dpi.

(C and D) Relative accumulation of PVY CP and ToBRFV CP transcripts in plants shown in (A) and (B), respectively.

(E and F) Relative accumulation of viral CP proteins in the PVY- and ToBRFV-infected plants shown in (A) and (B), respectively.

(G) Working model for COR15-mediated regulation of antiviral RNA silencing. During CTV infection, RNA silencing-mediated immunity is activated, and SGS3/RDR6 bodies, also known as siRNA bodies, mediate secondary vsiRNA synthesis and amplification of the RNA silencing response. The viral effector p20 targets SGS3 for autophagic degradation to suppress antiviral immunity. In turn, plants have evolved COR15 as a counter-counter-defense factor that interferes with p20-mediated autophagic degradation of SGS3 and promotes the formation of SGS3 condensates.

Discussion

Efficient siRNA processing and antiviral RNA silencing depend on SGS3 phase separation and the resultant assembly of siRNA bodies. This study demonstrates that the plant dehydrin COR15 enhances antiviral RNA silencing and suppresses CTV infection. Mechanistically, COR15 inhibits p20-mediated autophagic degradation of SGS3 by interfering with the p20–SGS3 interaction and promoting SGS3 condensate formation, thereby enhancing siRNA processing. Our findings elucidate the molecular mechanism by which plant secondary defense attenuates VSR activity and reveal a previously unknown function of dehydrins in the antiviral RNA silencing pathway.

Dehydrins are induced by diverse abiotic stresses and contribute to plant resistance to drought, cold, salinity, and heavy metals (Graether and Boddington, 2014). Recent findings have also shown that dehydrins participate in plant defense responses against fungal and bacterial pathogens (Rosales et al., 2014; Drira et al., 2016). Moreover, the expression of the dehydrin gene RAB25 is significantly increased during rice yellow mottle virus infection, suggesting that dehydrins may be involved in plant–virus interactions (Ventelon-Debout et al., 2004). Here, we found that NbCOR15 expression was significantly induced by CTV infection (Supplemental Figure 2), consistent with the responsiveness of dehydrins to both abiotic and biotic stresses. COR15 homologs from C. paradisi and C. unshiu have been shown to confer chilling tolerance in citrus after heat treatment (Porat et al., 2002; Hara et al., 2005). In this study, we found that transient overexpression of COR15 inhibited CTV infection, whereas COR15 silencing promoted CTV infection, in both N. benthamiana and citrus leaves (Supplemental Figures 3, 4, and 6). Consistently, the loss-of-function mutants cor15-4 and cor15-6 showed increased susceptibility to CTV infection, whereas heterologous expression of CaCOR15 in N. benthamiana enhanced plant resistance to CTV infection (Figure 2). Therefore, our results demonstrate that COR15 from the natural citrus host C. aurantifolia and the experimental host N. benthamiana possess antiviral activity, expanding our understanding of dehydrin function in plant antiviral immunity. In this study, the antiviral activity of CaCOR15 was evaluated through heterologous expression in N. benthamiana and transient expression in citrus leaves. A more direct approach, such as generating stable CaCOR15-overexpressing transgenic citrus plants followed by CTV inoculation, will be required to rigorously assess the antiviral role of CaCOR15 in its natural host.

Plant viruses have evolved counter-defense mechanisms, such as the expression of VSRs, to suppress antiviral RNA silencing and enable successful infection (Lopez-Gomollon and Baulcombe, 2022). In parallel, host plants have evolved counter-counter-defense strategies to antagonize VSR-mediated suppression of RNA silencing. For instance, the Tsw gene recognizes the tomato spotted wilt virus NSs protein and activates robust immunity (Margaria et al., 2007; Chen et al., 2023). Our previous study revealed that p20 mediates SGS3 degradation to suppress RNA silencing (Zhang et al., 2025). Here, we found that COR15 interacted with p20 and attenuated its VSR activity (Figure 3A–3D). Chen et al. (2020) and Jiang et al. (2021) showed that ubiquitin-like protein 5 and P3IP counteract the suppression of RNA-silencing by promoting the degradation of rice stripe virus P3 protein through the proteasome and autophagy pathways, respectively. The calmodulin-like protein rgs-CaM in tobacco inhibits RNA silencing suppression by promoting the autophagy-mediated degradation of several VSRs (Nakahara et al., 2012). However, our results showed that COR15 did not affect p20 protein accumulation (Figure 3D). Given that p20 exerts its VSR function by destabilizing SGS3, we analyzed the effect of COR15 on SGS3 protein accumulation. COR15 prevented p20-mediated SGS3 degradation in a dose-dependent manner (Figure 4A–4C). Similarly, AtHIRD11, a COR15 homolog in A. thaliana, protects its target protein from heavy metal-induced damage (Hara et al., 2016). This protective capacity may be attributable to the intrinsically disordered structures of dehydrins, which allow them to bind substrate molecules and protect them from stress-induced damage (Liu et al., 2017). Further mechanistic analysis revealed that COR15 bound to SGS3 and interfered with the p20–SGS3 interaction (Figure 4F–4K), thereby reducing the proportion of SGS3 localized to autophagosomes in the presence of p20 (Figure 4D). Moreover, both NbCOR15 and p20 associated with the N-terminus of NbSGS3, suggesting that COR15 may compete with p20 for the same N-terminal binding region to disrupt the p20–NbSGS3 interaction (Supplemental Figure 12). Future studies combining site-directed mutagenesis and structural analyses will be required to pinpoint the key residues responsible for this competitive binding and thereby clarify how NbCOR15 interferes with the p20–NbSGS3 interaction at the molecular level. Together, these findings uncover a previously unrecognized dehydrin-mediated counter-counter-defense mechanism in which COR15 directly antagonizes a viral suppressor of RNA silencing by stabilizing a host immune-related protein against VSR-mediated degradation.

Biomolecular condensates are membraneless organelles that are tightly regulated by developmental and environmental cues and often form through LLPS (Liu et al., 2024). Recent findings highlight the emerging roles of immune-related condensates in plants, including salicylic acid-induced NPR1 condensates, hematopoietic protein 1-driven condensates, and guanylate-binding protein-like 3-activated condensates (Zavaliev et al., 2020; Huang et al., 2021; Zhou et al., 2023). As a fundamental antiviral immune mechanism in eukaryotes, the RNA silencing pathway relies on the coordinated activity of multiple condensates distributed throughout the nucleus and cytoplasm, including dicing bodies, processing bodies, Cajal bodies, and siRNA bodies (Li et al., 2024). These condensates provide subcellular compartmentalization for sRNA biogenesis and activity and enable rapid responses to changes in the life cycle and environmental conditions (Solis-Miranda et al., 2023). SGS3 undergoes phase separation to drive siRNA body assembly and promote siRNA processing and antiviral immunity (Han et al., 2023). Although several regulatory factors involved in SGS3 condensate formation, such as SGS3-associated RNAs, translational inhibition, and mRNA decay, have been identified in recent studies (Han et al., 2023; Tan et al., 2023), our understanding of the molecular mechanisms and components underlying SGS3 condensate formation remains in its infancy, especially during viral infection. Our work revealed that CTV infection induces COR15, which interacts with SGS3 and localizes it to siRNA bodies (Figures 4F–4I and 5A). Moreover, we found that COR15 was incorporated into SGS3 condensates, which exhibited liquid-like properties, including movement and fusion into larger bodies (Figures 5B–5F), suggesting that COR15 is a novel component of siRNA bodies. The efficiency of siRNA processing is proportional to the abundance and proper state of siRNA bodies (Tan et al., 2023). Reported siRNA body components include DCLs, AGO1, SDE5, RH20, and the epigenetic factor Flowering locus VE (FVE), which coordinate siRNA production. In this process, DCLs help maintain the proper size of SGS3 condensates (Sun et al., 2021; Tan et al., 2023; Wen et al., 2024). We found that COR15 overexpression significantly increased the density of SGS3 condensates (Figures 5G and 5H). In contrast, mutations in NbCOR15 reduced the number of SGS3 condensates and increased the formation of large, irregular SGS3 aggregates (Figure 5I–5K), which resemble the dominant-negative SGS3dXS condensates (Tan et al., 2023). Consistently, COR15 promoted siRNA body formation and SGS3-dependent tasiRNA biogenesis (Figure 5L; Supplemental Figures 16 and 18). Moreover, COR15 homologs are widely distributed across plants, and multiple COR15 proteins enhanced RNA silencing in N. benthamiana (Supplemental Figure 8). These findings suggest that COR15 may be a conserved regulator of RNA silencing in plants. Future studies are needed to investigate how other SGS3 partners contribute to siRNA body assembly.

Dehydrins possess disordered features that favor phase separation, and certain LEA family proteins in A. thaliana and Artemia franciscana are induced to form condensates through LLPS during desiccation (Belott et al., 2020; Ginsawaeng et al., 2021). Here, we found that COR15 undergoes LLPS in siRNA bodies (Figures 5B–5F), further supporting the notion that dehydrins participate in biomolecular condensate formation. When expressed alone, COR15 was uniformly distributed in the cytoplasm and nucleoplasm (Supplemental Figure 1). However, upon CTV infection, p20 expression, or SGS3 overexpression, COR15 formed granules in the cytoplasm (Figures 1E, 1F, and 5A). We hypothesize that, when plants are subjected to viral infection or other stress conditions, certain dehydrins shift from a resting or dispersed state to an active condensate-associated state, thereby executing their functions. Future studies are required to test this possibility. Because dehydrins are involved in various stress responses in plants, dehydrin-mediated condensate formation may be a key mechanism that enables plants to respond rapidly to environmental changes and withstand stress. Therefore, determining how dehydrins participate in LLPS and promote biomolecular condensate formation will be crucial for deciphering the regulatory mechanisms of protein phase separation and understanding stress-response mechanisms in plants.

Based on these findings and our previous study (Zhang et al., 2025), we propose a working model illustrating a novel arms race mechanism in plant–virus interactions (Figure 7G). In CTV-infected cells, virus-derived dsRNAs are recognized and cleaved by DCLs into vsiRNAs, initiating RNA silencing-based antiviral defense. SGS3 drives the assembly of siRNA bodies to assist RDR6 in synthesizing additional dsRNAs, thereby amplifying antiviral RNA silencing. Concurrently, CTV expresses the VSR protein p20, which targets SGS3 for autophagic degradation to overcome plant defense. During ongoing plant–virus co-evolution, host plants have evolved a COR15-mediated mechanism to stabilize SGS3 by disrupting the p20–SGS3 interaction and to enhance SGS3 activity by promoting the formation of SGS3 condensates. This mechanism ultimately enhances siRNA processing and plant defense against CTV infection. Our results provide evidence for how plants subvert VSR-mediated counter-defense to suppress viral infection. Moreover, our findings elucidate a previously unknown role and mechanism by which a dehydrin protein exerts antiviral activity and identify potential targets for developing novel antiviral strategies.

Methods

Plant materials, Agrobacterium-mediated transient gene expression, and virus inoculation

N. benthamiana and C. sinensis plants used in this study were grown in a growth chamber maintained at 25°C under a 16-h light/8-h dark photoperiod and 60% relative humidity. The GFP-expressing transgenic N. benthamiana line 16c was described previously (Ruiz et al., 1998) and was provided by Prof. Xueping Zhou (Zhejiang University).

For transient expression in N. benthamiana, plant binary vectors were individually transformed into Agrobacterium strain GV3101. Agrobacterium cultures harboring the target plasmids were grown individually, collected, and resuspended in infiltration buffer containing 10 mM MgCl₂, 10 mM 2-(N-morpholino)ethanesulfonic acid (pH 5.6), and 150 μM acetosyringone. The resuspended Agrobacterium cultures were incubated at 28°C for 2–5 h and then infiltrated individually into leaves of 4-week-old N. benthamiana plants using needleless syringes. For transient expression in C. sinensis, plant binary vectors were individually transformed into Agrobacterium strain EHA105. Agrobacterium cultures were grown, collected, and resuspended in infiltration buffer to an OD₆₀₀ of 0.8–1.0. After incubation at 28°C for 2–5 h, the resuspended Agrobacterium cultures harboring the recombinant plasmids were infiltrated into CTV-infected C. sinensis leaves as described previously (Long et al., 2021). The infiltrated leaves were collected at 5 dpi for RT–qPCR and immunoblot analyses.

For CTV inoculation, Agrobacterium cultures containing the CTV infectious clone were grown, collected, diluted to an OD₆₀₀ of 0.2 in infiltration buffer, and incubated at 28°C for 2–5 h. The cultures were then infiltrated into the leaves of 4-week-old or VIGS-treated N. benthamiana plants. Plants infiltrated with infiltration buffer served as negative controls. The experiment was repeated three times with 15–20 plants per treatment.

For PVY-GFP and ToBRFV-GFP inoculation, Agrobacterium cultures harboring the corresponding infectious clones were grown, harvested by centrifugation, and resuspended to an OD₆₀₀ of 0.5 in infiltration buffer. The cultures were incubated at 28°C for 2–5 h before infiltration into leaves of 4-week-old N. benthamiana plants. Each experiment was repeated three times, with 5–10 plants per treatment.

Plasmid construction

The full-length coding sequences of CaCOR15, NbCOR15, NbSGS3, β-glucuronidase (GUS), and CTV p20 were amplified and individually cloned into the pPR3-N, pDHB1, pGBKT7, pJW771-nLUC, pJW772-cLUC, pCNF3, pCNF3-YFP-FLAG, and pCNF3-CFP-HA vectors to generate NubG-CaCOR15, NubG-NbCOR15, NbSGS3-Cub, p20-Cub, BD-p20, NbCOR15-nLUC, cLUC-CaCOR15, cLUC-NbCOR15, cLUC-NbSGS3, GUS-nLUC, cLUC-GUS, CaCOR15-FLAG, NbCOR15-FLAG, NbCOR15-HA, NbCOR15-Myc, CaCOR15-YFP-FLAG, NbCOR15-YFP-FLAG, CaCOR15-CFP-HA, and p20-CFP-HA constructs using the ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China) according to the manufacturer’s instructions. The full-length coding sequences of CaCOR15 and NbCOR15 were individually cloned into the pEarleyGate202-YC vector to generate CaCOR15-Y^C and NbCOR15-Y^C constructs using Gateway technology (Invitrogen, Carlsbad, CA, USA). To generate the proSGS3:NbSGS3-CFP plasmid, the 2000-bp promoter region upstream of NbSGS3 was amplified from N. benthamiana genomic DNA and subsequently cloned into the NbSGS3-CFP-HA vector, as described previously (Zhang et al., 2025). For VIGS assays, the SGN-VIGS website (https://vigs.solgenomics.net/) was used to predict 300-bp specific target sequences for silencing NbCOR15, NbSGS3, and CsCOR15. The target fragments were amplified and individually inserted into the pTRV2 vector to generate pTRV2-NbCOR15, pTRV2-NbSGS3, and pTRV2-CsCOR15. To generate NbCOR15-knockout N. benthamiana plants, a single-guide RNA (sgRNA) sequence (5′-GCAGCGGAGAGCACAAGGAAGGG-3′) was designed using CRISPR-P 2.0 (http://cbi.hzau.edu.cn/CRISPR2/). A tandemly arrayed tRNA–gRNA cassette containing a single sgRNA was recombined into the pRGEB35 vector under the control of the A. thaliana ubiquitin promoter as described previously (Xie et al., 2015). The plasmids p20-nLUC, NbSGS3-Y^N, NbRDR6-Y^C, p20-Y^N, GST-Y^N, GST-Y^C, NbSGS3-CFP-HA, YFP-NbSGS3, NbRDR6-HA, p20-FLAG, and p20-YFP-FLAG were constructed as described previously (Zhang et al., 2025). The primers used for plasmid construction are listed in Supplemental Table 1.

Generation of transgenic N. benthamiana plants

To generate NbCOR15-knockout and CaCOR15-expressing transgenic N. benthamiana plants, the pRGEB35-NbCOR15 and CaCOR15-FLAG plasmids were individually transformed into WT N. benthamiana plants via Agrobacterium-mediated leaf disc transformation (BioRun, Wuhan, China). The NbCOR15-knockout plants were validated by amplifying and sequencing DNA fragments spanning the target sites from the genomic DNA of transformed plants. CaCOR15 expression was verified by amplifying and sequencing a CaCOR15-specific fragment from the cDNA of transgenic plants. The transgenic N. benthamiana lines were grown in a growth chamber to obtain T2-generation lines, which were identified by PCR and sequencing and used for subsequent experiments. The primers used for these assays are listed in Supplemental Table 1.

Nucleic acid extraction, RT–qPCR, and northern blotting assays

Total DNA was extracted using the Plant gDNA Extract Kit (Coolaber, Beijing, China). Total RNA was extracted using TRIzol reagent (Invitrogen). Small RNAs for northern blotting assays were extracted using the miRcute miRNA Isolation Kit (Tiangen, Beijing, China). First-strand cDNA was synthesized using the HiScript II Q Select RT SuperMix for qPCR Kit (Vazyme, Nanjing, China). tasiR-ARF3 levels were quantified by stem-loop RT–qPCR as described previously (Zhang et al., 2025). All qPCR assays were performed using ChamQ SYBR Color qPCR Master Mix (Vazyme, Nanjing, China) on a CFX96 instrument (Bio-Rad, Hercules, CA, USA). The N. benthamiana Actin gene was used as an internal control. Transcript levels of the analyzed genes were calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001). The DIG Northern Starter Kit (Roche, Basel, Switzerland) was used for northern blotting assays. For GFP mRNA blotting, a GFP fragment was labeled with digoxigenin (DIG) and used as a probe, and 1 μg of total RNA was separated on 1.2% agarose gels containing 2% formaldehyde, followed by transfer onto Hybond-N⁺ membranes (GE Healthcare, Chicago, IL, USA). After UV crosslinking, the membranes were hybridized with DIG-labeled probes at 68°C. For siRNA detection, probes were synthesized and DIG-labeled by Sangon (Shanghai, China). The sRNA extracts were separated by 15% PAGE and transferred onto Hybond-N⁺ membranes. The membranes were UV-crosslinked and hybridized with siRNA probes at 42°C. The hybridized probes were detected and visualized using an anti-DIG-AP antibody and CDP-Star substrate with the ChemiDoc Touch Imaging System (Bio-Rad). The primers and probes used for qPCR and northern blotting assays are listed in Supplemental Table 1 or were described in our previous study (Zhang et al., 2025).

VIGS assay

TRV-mediated VIGS in N. benthamiana was performed as described previously (Liu et al., 2002). Agrobacterium cultures (OD₆₀₀ = 0.6) harboring pTRV1 were mixed with an equal volume of Agrobacterium cultures (OD₆₀₀ = 0.6) harboring pTRV2 (TRV:00), pTRV2-NbCOR15 (TRV:NbCOR15), or pTRV2-NbSGS3 (TRV:NbSGS3). The mixed cultures were individually infiltrated into the leaves of 4-week-old N. benthamiana plants. Gene-silencing efficiency was determined by RT–qPCR at 10 dpi.

For transient VIGS assays in C. sinensis leaves, Agrobacterium cultures (OD₆₀₀ = 0.6) harboring pTRV1 and either pTRV2 (TRV:00) or pTRV2-CsCOR15 (TRV:CsCOR15) were mixed at a 1:1 volume ratio and infiltrated into the leaves of CTV-infected C. sinensis plants as described above. Gene-silencing efficiency and virus accumulation were determined by RT–qPCR at 7 dpi.

Y2H assay

Y2H screening assays based on the GAL4 Y2H system were conducted using the Make Your Own “Mate & Plate” Library System (Clontech, San Francisco, CA, USA) according to the manufacturer’s protocols. The p20 gene was cloned into the pGBKT7 vector as bait. A Y2H prey library constructed from C. aurantifolia cDNA was used to screen for p20-interacting proteins. Directed Y2H assays were performed using the split-ubiquitin Y2H system (Dualsystems, Schlieren, Switzerland) according to the manufacturer’s protocols. Different combinations of bait and prey plasmids were transformed into the yeast strain NMY51 (TaKaRa, Beijing, China). The transformed yeast cells were grown on double-dropout medium (SD/−Trp−Leu) for 2–3 days at 28°C and then spotted onto quadruple-dropout medium (SD/−Trp−Leu−His−Ade) for 3–4 days at 28°C. The plasmid pair pDHB1-largeT/pDSL-p53 served as a positive control, and the plasmid pair pDHB1-largeT/pPR3-Ost1 served as a negative control.

BiFC and subcellular localization assays

BiFC and subcellular localization assays were performed as described previously (Zhang et al., 2025). Agrobacterium cultures containing different combinations of constructs were infiltrated into N. benthamiana leaves. Fluorescence in the infiltrated leaves was observed using a Leica TCS-SP8 confocal microscope (Wetzlar, Germany) at 2–3 dpi. CFP was excited at 458 nm and detected at 470–510 nm. YFP was excited at 514 nm and detected at 525–565 nm. mCherry was excited at 561 nm and detected at 585–625 nm.

LCI and competitive LCI assays

LCI assays were performed as described previously (Zhou et al., 2018). Different combinations of Agrobacterium mixtures (OD₆₀₀ = 0.5) were infiltrated into separate areas of the same N. benthamiana leaf. The GUS-nLUC and cLUC-GUS vectors were used as negative controls. The infiltrated leaves were sprayed with 0.2 mM D-luciferin (GlpBio, Montclair, NJ, USA) for 10–15 min at 2 dpi. Luciferase activity was assessed using a low-light cooled charge-coupled device imaging apparatus (NightSHADE L985; Berthold, Stuttgart, Germany). For the competitive LCI assay, p20-nLUC and cLUC-NbSGS3 were co-infiltrated into N. benthamiana leaves along with different dilutions of NbCOR15-FLAG or an empty vector. Luciferase activity was assessed at 2 dpi.

Plant protein extraction and immunoblotting

N. benthamiana leaves were collected, ground in liquid nitrogen, and mixed with two volumes of ice-cold protein extraction buffer (25 mM Tris [pH 7.5], 10% glycerol, 1 mM EDTA, 150 mM NaCl, 2% [w/v] polyvinylpyrrolidone, 10 mM dithiothreitol, 0.1% NP-40, and 1× protease inhibitor cocktail). After lysis on ice for 10–30 min, the samples were centrifuged at 12,000 rpm for 15 min at 4°C. The supernatants were collected, mixed with 2× SDS loading buffer, and boiled for 10 min. For immunoblotting, proteins were separated by SDS–PAGE and transferred onto polyvinylidene fluoride membranes (Millipore, Billerica, MA, USA). Proteins were then detected by immunoblotting using anti-FLAG, anti-HA, anti-Myc, anti-GFP (ABclonal, Wuhan, China), or anti-CTV CP (Wang et al., 2006) antibodies, followed by horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies (ABclonal, Wuhan, China). Finally, the membranes were washed, incubated with WesternBright ECL HRP substrate (Advansta, San Jose, CA, USA), and visualized using the ChemiDoc Touch Imaging System (Bio-Rad).

Co-IP and competitive co-IP assays

For co-IP assays, protein extracts were incubated with anti-FLAG beads (Sigma-Aldrich, St. Louis, MO, USA) at 4°C for 2 h. The beads were washed five times with IP buffer (25 mM Tris [pH 7.5], 10% glycerol, 1 mM EDTA, 150 mM NaCl, 0.1% NP-40, and 1× protease inhibitor cocktail) and boiled for 10 min in 2× SDS loading buffer. For competitive co-IP assays, p20-FLAG and YFP-NbSGS3 were co-infiltrated into N. benthamiana leaves together with NbCOR15-HA or CFP-HA. Total protein was extracted at 2 dpi and immunoprecipitated with anti-FLAG beads, followed by immunoblotting using anti-FLAG, anti-GFP, and anti-HA antibodies.

FRAP assay

FRAP assays were performed using a Leica TCS-SP8 confocal microscope. Target regions were bleached with a 514-nm laser at 100% power until the fluorescence intensity of the selected areas was reduced to less than 20% of the pre-bleaching intensity. Fluorescence recovery was recorded at 2-s intervals for 120 s after bleaching. Fluorescence intensity at each time point was measured using the microscope software.

Chemical treatment and MDC staining for autophagosome observation

For 1,6-hexanediol treatment (Macklin, Shanghai, China), N. benthamiana leaves were infiltrated with 10% 1,6-hexanediol for 5–15 min before imaging, and H₂O treatment was used as a negative control. To activate autophagy, leaves were treated with 100 μM BTH (Sigma-Aldrich) for 12–16 h, with DMSO treatment serving as a negative control.

For MDC staining and autophagosome observation, infiltrated N. benthamiana leaves were treated with 100 μM E-64d (Sigma-Aldrich) at 36 hpi and placed in the dark for 12 h. The leaves were then vacuum-infiltrated with 50 μM MDC (Sigma-Aldrich) for 10 min. MDC-stained structures were excited at 405 nm and detected at 410–585 nm. Chlorophyll autofluorescence was detected at 690–730 nm.

Protein stability assay

For the NbSGS3 stability assay, p20-FLAG and YFP-NbSGS3 were co-infiltrated with EV or NbCOR15-MYC into N. benthamiana leaves. The infiltrated leaves were treated with 10 mM 3-MA (Sigma-Aldrich), 100 μM MG132 (Sigma-Aldrich), or an equivalent volume of DMSO as a control. Total protein was extracted at 12 h after chemical treatment and incubated at 25°C with 100 μM cycloheximide (Sigma-Aldrich) and 10 mM ATP. Samples were collected at the designated time points for analysis.

Statistical analysis

Data are presented as means ± standard deviation (SD) from three independent experiments. Statistical significance was assessed using a one-tailed Student’s t-test or one-way ANOVA with Tukey’s multiple-comparison test. Immunoblot bands were quantified using ImageJ software and normalized against the loading controls. ImageJ was used to measure the number and size of SGS3 condensates.

Data and code availability

All data supporting the findings of this study are included in the main article and the supplemental information. Gene sequence data used in this study are available in the GenBank database under the following accession numbers: CaCOR15 (GenBank: OQ706358.1), NbCOR15 (GenBank: OQ706359.1), NbSGS3 (GenBank: KJ190939.1), NbRDR6 (GenBank: AY722008.1), NbActin (GenBank: JQ256516.1), and p20 (GenBank: OQ708955.1).

Funding

This research was supported by grants from the National Natural Science Foundation of China (nos. 32502452 and 31870145) and the Taishan Scholar Construction Project (no. ts2022028).

Acknowledgments

We thank Prof. Tao Zhou (China Agricultural University) for helpful suggestions. We are grateful to Dr. Yan Zhou (Southwest University), Prof. Kabin Xie (Huazhong Agricultural University), and Prof. Xiaoli Guo (Huazhong Agricultural University) for kindly providing the CTV infectious clone and the pRGEB35, pJW771-nLUC, and pJW772-cLUC vectors. No conflict of interest declared.

Author contributions

X.-D.L., N.H., G.-P.W., Y.-P.T., and Y.-L.Z. conceived the project and designed the experiments. Y.-L.Z., Z.-K.Y., Z.Z., L.-Y.M., and S.-S.J. performed the experiments. All authors analyzed the data. X.-D.L., N.H., and Y.-L.Z. wrote the original manuscript. X.-D.L. and Y.-L.Z. revised the manuscript.

Published: April 30, 2026