Pepper mild mottle virus coat protein interacts with pepper chloroplast outer envelope membrane protein OMP24 to inhibit antiviral immunity in plants

Abstract Pepper mild mottle virus (PMMoV) is a devastating viral pathogen of pepper (Capsicum annuum) but it is unclear whether and how peppers protect against PMMoV infection. The expression of the chloroplast outer membrane protein 24 (OMP24) of C. annuum was upregulated under PMMoV infection and it interacted with PMMoV coat protein (CP). Silencing of OMP24 in either C. annuum or Nicotiana benthamiana facilitated PMMoV infection, whereas overexpression of N. benthamiana OMP24 in transgenic plants inhibited PMMoV infection. Both C. annuum OMP24 (CaOMP24) and N. benthamiana OMP24 (NbOMP24) localized to the chloroplast and have a moderately hydrophobic transmembrane domain that is necessary for their localization. Overexpression of CaOMP24 induced stromules, perinuclear chloroplast clustering, and accumulation of reactive oxygen species (ROS), the typical defense responses of chloroplasts transferring the retrograde signaling to the nucleus to regulate resistance genes. The expression of PR1 and PR2 was also upregulated significantly in plants overexpressing OMP24. Self-interaction of OMP24 was demonstrated and was required for OMP24-mediated plant defense. Interaction with PMMoV CP interfered with the self-interaction of OMP24 and impaired OMP24-induced stromules, perinuclear chloroplast clustering and ROS accumulation. The results demonstrate the defense function of OMP24 in pepper during viral infection and suggest a possible mechanism by which PMMoV CP modulates the plant defense to facilitate viral infection.


Introduction
Pepper mild mottle virus (PMMoV), a member of the genus Tobamovirus, is the major viral pathogen of pepper (Capsicum annuum L.), resulting in significant crop losses around the world [1,48]. PMMoV is transmitted directly to plants by infected sap and also through contaminated seed. PMMoV has a singlestranded RNA genome (about 6.4 Kb) that encodes four proteins: p126 (viral suppressor of RNA silencing), p183 (RNA-dependent RNA polymerase), movement protein, and coat protein (CP) [4]. The CP of all tobamoviruses is encoded by a subgenomic RNA and plays a versatile role during the viral infection process. Besides encapsidating the viral genome, it is involved in systemic virus movement, cross-protection, pathogenicity, and symptom development [55]. Changes in the PMMoV CP amino acid sequence are reported to be responsible for breaking the L 3 and L 4 mediatedresistance of the host [6,24]. Our previous study showed that the 20th amino acid of PMMoV CP determined the chlorosis symptom and is responsible for subcellular localization to the chloroplast [25]. However, the detailed biological function of the relationship between PMMoV CP and chloroplasts has not been determined.
In addition to supplying the plant with energy, chloroplasts are the source of defense signals including phytohormones, reactive oxygen species (ROS) and calcium (Ca 2+ ), hence playing critical roles in plant immunity [36,42]. Moreover, as an environmental sensor, chloroplasts communicate with the nucleus to change the expression of thousands of proteins, which is termed retrograde signaling [15]. Perinuclear chloroplast clustering (PCC) and the formation of stromules that are dynamic tubular extensions from chloroplasts are general plant immune responses of plants under stresses [13,16,17,42]. The movement of chloroplasts to the nucleus is believed to be guided by stromules. A retrograde signal such as ROS originating in the chloroplasts is transferred to the nucleus through the stromules connecting the chloroplast to the nucleus, regulating resistance gene expression and thereby mediating plant immunity [26,27].
Chloroplasts have a double-membrane structure. The outer envelope membrane is a biochemical and physical barrier and the outer membrane proteins (OMPs) play important roles in intracellular communication and organelle biogenesis [5,21,28,29,31]. In addition, chloroplast outer membrane proteins DGD1, EDS5, and JASSY have been shown to participate in the biosynthesis of the lipid digalactosyldiacylglycerol (DGDG), SA and jasmonate, respectively [23,41,56]. It is not known if and how the OMPs participate in plant defense.

Figure 1.
PMMoV CP interacts with CaOMP24. A PMMoV CP interacted with CaOMP24 in a yeast two hybrid (Y2H) assay. Yeast NMY51 co-transformed with pDHB1-large T + pDSL-p53, pDHB1-PMMoV CP + pOst1-NubI was the positive control; pDHB1-large T + pPR3-N, pDHB1-PMMoV CP + pPR3-N were the negative controls. Serial 10-fold dilutions of yeast cells were plated on selective medium (SD)/−Trp, −Leu, -His, −Ade or (SD)/−Trp, −Leu for 3-5 days. B Split-luciferase complementation (LCI) assays demonstrating that PMMoV CP interacts with CaOMP24 in N. benthamiana leaves. The N-or C-terminal fragments of luciferase (LUC) were fused to the C-terminus of PMMoV CP and N-terminus of CaOMP24. β-glucuronidase (GUS) was used as a negative control. C Co-IP analysis of the interaction between PMMoV CP and CaOMP24. N. benthamiana leaves transiently co-expressing CaOMP24-GFP with PMMoV CP-Myc or pGUS-Myc were harvested at 60 hpi. Total proteins were immunoprecipitated with anti-GFP beads and samples before (Input) and after (IP) immunopurification were detected by western blot with an anti-GFP or anti-Myc antibody. The red asterisks indicate the expected band sizes. D BiFC analysis of the interaction between PMMoV CP and CaOMP24. The N-(nYFP) or C-terminal (cYFP) fragments of YFP were fused to the C-terminus of PMMoV CP and CaOMP24. pGUS was used as a negative control. Confocal analysis was performed at 2 dpi. Scale bars, 20 μm. E Summary of interactions between PMMoV CP and the CaOMP24 truncated mutants as determined by Y2H and LIC.
Here, we identified that the chloroplast OMP24 has a defense role in pepper during PMMoV infection. Overexpression of OMP24 caused stromules, PCC, and accumulation of ROS, as well as the upregulated expression of resistance genes. PMMoV CP interacts with OMP24 and interferes with the self-interaction of OMP24 and hence inhibits the OMP24-mediated resistance to facilitate PMMoV infection. These results provide the first evidence for a mechanism by which OMPs are involved in plant defense against viral infection.

PMMoV CP interacts with CaOMP24
To identify possible host proteins required for PMMoV CP to perform its function in viral infection, a cDNA library of C. annuum was used for protein screening with PMMoV CP as the bait by a split-ubiquitin membrane yeast two-hybrid DUALhunter system. One of the candidate proteins gave a positive signal on the selective media (SD/−Trp-Leu-His-Trp) (Fig. 1A). The prey fragment in this possible interaction was sequenced with primers CYC1-F/R. Result showed that it covered 84% of the Capana12g002319 CDS in the C. annuum database (https://solgenomics.net/ftp/genomes/ Capsicum_annuum/). Capana12g002319 contains an ORF of 450 nucleotides predicted to encode a 149-amino acid protein that has 30.2% identity to chloroplast OMP24 in spinach (Spinacia oleracea L.) (Fig. S1, see online supplementary material). We therefore named Capana12g002319 as CaOMP24. OMP24 in spinach is reported to be an acidic protein (calculated isoelectric point 4.8) deeply embedded in the chloroplast outer membrane [18]. Alignment of protein sequences from C. annuum, Nicotiana benthamiana, Arabidopsis thaliana and S. oleracea showed that OMP24 has a conserved motif near its C-terminus (Fig. S1, see online supplementary material).
To analyse the key region of CaOMP24 that interacted with PMMoV CP, we divided CaOMP24 into two peptides (N-terminal N80 and C-terminal N80). Both Y2H and LCI experiments confirmed that CaOMP24 N80 but not CaOMP24 N80 could interact with PMMoV CP (Fig. S2B and C, see online supplementary material). To better define the region of CaOMP24 required for the interaction, we used a series of truncated mutants of CaOMP24 with 1st-28th, 29th-49th or 50th-80th amino acids for analysis. The mutant CaOMP24 1-28 interacted with PMMoV CP, but there was no interaction with either CaOMP24  or CaOMP24 50-80 ( Fig. S2B and C, see online supplementary material ), indicating that the N-terminal 28 amino acid region of CaOMP24 is necessary for its interaction with PMMoV CP.

Silencing of OMP24 facilitates PMMoV infection in pepper and N. benthamiana
To investigate the function of CaOMP24 in pepper plants during PMMoV infection, the tobacco rattle virus (TRV)-induced gene silencing (VIGS) system was used to silence CaOMP24 and the plants were then inoculated with PMMoV. For analysis, a partial sequence of CaOMP24 was inserted into RNA2 of TRV, producing TRV:CaOMP24. At 18 dpi of TRV:CaOMP24 inoculation, the expression of CaOMP24 was decreased to ∼30% of the normal level in the control inoculated with the empty TRV (TRV:00), while the plants did not show any obvious phenotype ( Fig. 2A; Fig. S3A, see online supplementary material). Plants were then rub-inoculated with PMMoV. At 30 dpi of PMMoV infection, the silenced plants Total protein was extracted from PMMoV-GFP-inoculated leaves at 4 dpi or systemically infected leaves at 7 dpi. had more severe disease symptoms than the control with more curling and shrinking leaves ( Fig. 2A). Consistently, PMMoV CP accumulated at a higher level in the silenced plants (Fig. 2B).
We also did similar experiments in N. benthamiana, a model plant for studying virus-host interactions [22]. There are two CaOMP24 homologues in N. benthamiana with 96.7% nucleotide identity and 95.0% amino acid identity, namely NbOMP24.1 (Niben101Scf02622g10012) and NbOMP24.2 (Niben101Scf05697g03010). Both NbOMP24.1 and NbOMP24.2 interacted with PMMoV CP in Y2H, LIC, and BIFC experiments (Fig. S4, see online supplementary material). We investigated the role of NbOMP24s in PMMoV infection using the TRV-VIGS system. Due to their high sequence identity, NbOMP24.1 and NbOMP24.2 would be silenced simultaneously in VIGS analysis. Silencing of NbOMP24s decreased the transcript level of NbOMP24s to 20% at 12 dpi but did not cause obvious developmental change in the plants ( Fig. S3B and C, see online supplementary material). Plants were then rub-inoculated with PMMoV-GFP. Compared to TRV:00treated plants, more f luorescent foci appeared on the inoculated leaves of NbOMP24-silenced plants at 4 dpi of PMMoV infection ( Fig. 2C and D). Moreover, the speed of systemic infection was remarkably faster in NbOMP24s-silenced plants than in the nonsilenced ones (Fig. 2E). The accumulation of PMMoV CP in both inoculated leaves (IL) and systemically infected leaves (SL) of silenced plants was significantly higher than that in the nonsilenced controls (Fig. 2F).
To eliminate any effect of TRV on PMMoV infection, transient gene silencing experiments were done to confirm the effect of silencing NbOMP24s on PMMoV infection. The left and right sides of a single N. benthamiana leaf were infiltrated with agrobacterium strain GV3101 containing the construct expressing hairpin RNA of NbOMP24 (hpNbOMP24) and hpGUS, respectively, and then inoculated with PMMoV-GFP. Results of qRT-PCR showed that the expression level of NbOMP24s was reduced to 28% of the control ( Taken together, these results demonstrate that silencing of OMP24 facilitated PMMoV infection in both pepper and N. benthamiana.

Overexpression of NbOMP24.1 suppresses PMMoV infection on N. benthamiana
The expression of CaOMP24 and NbOMP24s was up-regulated in plants during PMMoV infection ( Fig. S6A and B, see online supplementary material) but expression of CP alone did not affect the expression of NbOMP24s (Fig. S6C, see online supplementary material). We also detected the response of NbOMP24s to seven other pepper-infecting viruses. Three of the seven viruses tested caused an up-regulation in the expression of NbOMP24s, while the remaining four viruses had no effect on the expression of NbOMP24s (Fig. S6D, see online supplementary material), which suggests that there may be some specificity in the interactions between the viruses and the plant host.
To further confirm the roles of OMP24 in defense against PMMoV, transgenic N. benthamiana plants expressing Myc-tagged NbOMP24.1 driven by the caulif lower mosaic virus (CaMV) 35S promotor were generated. The expression of NbOMP24.1-myc in transgenic plants was confirmed by western blot, and the transgenic plants had a similar phenotype to wild-type (Fig. S7, see online supplementary material). Two independent transgenic lines (designated as OE#2 and OE#9) were chosen for viral challenge. At 4 dpi of PMMoV-GFP infection, the numbers of infection foci on the inoculated leaves of OE#2 and OE#9 plants were less than on the wild-type plants (Fig. 3A). At 7 dpi, the f luorescence associated with systemic infection of transgenic plants was extensive (Fig. 3A). The accumulation of PMMoV CP was less in both the inoculated and systemic leaves of transgenic plants than in control plants (Fig. 3B). The results further demonstrate the defense role of OMP24 against PMMoV infection.

CaOMP24 localizes to the chloroplast and induces stromules and PCC
To understand the possible mechanism by which OMP24 defends against PMMoV infection, we first investigated its chloroplast localization and its effect on chloroplasts in epidermal cells of N. benthamiana. The complete open reading frame of CaOMP24 was cloned into the pCV-N-GFP vector (CaOMP24-GFP) to express CaOMP24 with a C-terminal GFP tag. CaOMP24-GFP was then expressed in N. benthamiana leaves by agroinfiltration. At 24-48 hours post inoculation (hpi), green f luorescence was distributed on the chloroplast outer membrane (Fig. 4A and B). Additionally, expression of CaOMP24-GFP induced stromules on the chloroplasts (Fig. 4B). The green f luorescence from CaOMP24-GFP merged with the red f luorescence from the transit peptide of ferredoxin NADP(H) oxidoreductase (FNR-mCherry), a marker of stromules (Fig. 5). Notably, the chloroplasts progressively moved towards the nucleus when CaOMP24-GFP was expressed, finally being clustered around the nucleus (Fig. 4B).
CaOMP24 is a predicted to be a membrane protein.  localized to the cell periphery and nuclear envelope but did not induce perinuclear chloroplast clustering (PCC) (Fig. 4A-B). When only the TMD was fused with GFP, the f luorescence was localized to the chloroplast outer membrane as well as the cell periphery and nuclear envelope but did not induce either stromules or PCC (Fig. 4B). In the TMD, the phenylalanine (F) at residue 118 of CaOMP24 is a hydrophobic amino acid. If it was substituted with the hydrophilic amino acid, Lysine (K), the TMD is predicted to disappear (Fig. S8B, see online supplementary material). As expected, CaOMP24 F118K -GFP localized to the cell periphery and nuclear envelope but did not induce stromules or PCC (Fig. 4B). Moreover, expression of CaOMP24 mutants (CaOMP24 TMD , CaOMP24 F118K , CaOMP24 TMD ) resulted in significantly lower stromule induction compared to CaOMP24 expression (Fig. 5).
The expression of CaOMP24 and its mutants were confirmed using anti-GFP antibody (Fig. S8C, see online supplementary material). NbOMP24.1 has the same characteristics as CaOMP24. NbOMP24.1 was also localized to the chloroplast outer membrane and induced chloroplast clustering around the nucleus, while expression of mutants with deletion of the TMD (NbOMP24.1 TMD ) or substitution of phenylalanine for lysine (NbOMP24.1 F109K ) led to localization to the plasma membrane and did not cause chloroplast clustering (Fig. S9, see online supplementary material). Additionally, PMMoV infection induced stromules and PCC (Fig. S10, see online supplementary material). Taken together, these results demonstrate that the TMD, and the phenylalanine (F) residue in the TMD, are vital for OMP24 chloroplast localization and induction of stromules and PCC.

ROS accumulated in cells expressing CaOMP24 but not CaOMP24 F118K
Increasing evidence indicates that perinuclear chloroplast clustering (PCC) plays a fundamental defense role during plantpathogen interactions [13,14,16,38]. We therefore supposed that CaOMP24 might play its defensive role by inducing PCC. To investigate this, we first tested whether the CaOMP24 mutant that did not induce PCC was still able to defend against PMMoV. CaOMP24-Myc or CaOMP24 F118K -Myc were co-infiltrated with PMMoV-GFP into N. benthamiana leaves by agroinfiltration. At 4 dpi, the intensity of f luorescence in zones expressing CaOMP24-Myc was less than that in control zones expressing 00-Myc (empty vector), confirming that CaOMP24 expression decreased PMMoV-GFP infection (Fig. 6A). However, the intensity of f luorescence in zones expressing CaOMP24 F118K -Myc was similar to that in the control (Fig. 6A). Consistently, PMMoV CP accumulated at lower levels in the zones expressing CaOMP24 than in those expressing CaOMP24 F118K -Myc or the control (Fig. 6B). These results suggest that the induction of PCC is essential for the defense function of CaOMP24.
It is hypothesized that PCC helps to transfer the retrograde signal ROS into the nucleus to regulate immune responses via stromules [26]. Consistent with the results obtained with CaOMP24-GFP (Figs 4B and 5), the frequency of stromule formation connecting chloroplasts and nuclei in CaOMP24-Myc was higher than that of the control and CaOMP24 F118K -Myc (Fig. 6C). Furthermore, we monitored the total cellular ROS accumulation in cells expressing CaOMP24 and its mutants using 2 ,7 -dichlorodihydrof luorescein diacetate (DCFH-DA) staining. The accumulation of ROS in cells expressing CaOMP24 was dramatically increased compared to the control or cells expressing CaOMP24 F118K (Fig. 6D). Consistent with the DCFH-DA staining results, H 2 O 2 content in leaves expressing CaOMP24 was significantly higher than in leaves expressing 00myc, while this increase was compromised upon PMMoV infection (Fig. S11, see online supplementary material). Moreover, stromules and PCC were also significantly induced in NbOMP24.1transgenic plants (OE#2 and OE#9), accompanied by an accumulation of ROS ( Fig. S12A-C, see online supplementary material). As expected, the expression of PR1 and PR2, the resistant genes in ROS-mediated defense, was upregulated significantly in NbOMP24.1-transgenic plants (Fig. S12D, see online supplementary material). These results suggest that CaOMP24 may resist viral infection by the PCC-associated ROS-mediated defense pathway.
ROS are important signals mediating general resistance against pathogens [9,51]. We here also investigated whether NbOMP24 functions in defense against two other viruses, potato virus X (PVX, genus Potexvirus) and turnip mosaic virus (TuMV, genus

PMMoV CP impairs CaOMP24-induced PCC and accumulation of ROS by interfering with the CaOMP24 self-interaction
Because PMMoV CP interacted with CaOMP24, we next investigated the effect of PMMoV CP on the ability of CaOMP24 to induce PCC and ROS accumulation. Confocal analysis at 48 hpi revealed that the extent of PCC and stromule induction was less in leaves co-expressing CaOMP24-myc with PMMoV CP-Flag than in the control (CaOMP24-myc + pGUS-Flag) (Fig. 7A). Meanwhile, cellular ROS accumulation induced by CaOMP24 was obviously decreased in the presence of PMMoV CP (Fig. 7B). In addition, the mRNA and protein expression levels of CaOMP24 were not affected by co-expression with PMMoV CP (Fig. S15, see online supplementary material). These results indicate that PMMoV CP inhibits the antiviral function of CaOMP24.
Next, we wanted to know how CP impaired CaOMP24-induced PCC and accumulation of ROS. Because we had determined (above) that the TMD of CaOMP24 was important for its localization and induction of PCC and that the N terminal 1-28aa of CaOMP24 was the key domain for its interaction with PMMoV CP, we tested whether the N terminus 1-28aa of CaOMP24 was also essential for inducing stromules, PCC, and ROS accumulation. GFP-fused CaOMP24 N28 localized at chloroplast outer membranes but did not induce stromules, PCC, or ROS accumulation (Fig. 8A-C). In addition, overexpression of CaOMP24 N28 did not confer resistance to PMMoV (Fig. 8D). These results indicate that the N terminal 1-28aa of CaOMP24 are also essential for CaOMP24-induced stromules, PCC, and defense.
Several chloroplast OMPs including Toc33 and Toc159 have been reported to require self-interaction to perform their functions [7,57]. To analyse whether CaOMP24 functioned by interacting with itself, we performed Y2H, Co-IP, and LCI experiments to detect its self-interaction. CaOMP24 indeed interacted with itself ( Fig. 8E; Fig. S16, see online supplementary material) and the key region for its self-interaction was in the N terminal 1-80 aa. Although CaOMP24 interacted with all three truncated benthamiana leaf 48 hours after agroinfiltration examined under the confocal microscope. Scale bars, 20 μm. B Co-expression of CaOMP24 N28 -GFP with FNR-mCherry. The frequency of stromule formation was lower when CaOMP24 N28 was expressed compared to that of CaOMP24. Scale bars, 20 μm. Triangle, stromule. C Cellular ROS accumulation in cells expressing CaOMP24 N28 -myc was measured by DCFH-DA staining. Scale bars, 20 μm. D CaOMP24 N28 did not confer resistance to PMMoV. GFP f luorescence and PMMoV CP accumulation in the regions expressing either CaOMP24 or CaOMP24 N28 with PMMoV-GFP at 4 dpi. E Co-IP assay confirmed the self-interaction of CaOMP24. The red asterisks indicate the expected band sizes. F Competitive Co-IP assays demonstrating that the self-interaction of CaOMP24 was disrupted by PMMoV CP. The combination of CaOMP24-GFP + CaOMP24-myc was co-expressed with PMMoV CP-Flag or pGUS-Flag in N. benthamiana leaves. Total protein was extracted with GTEN buffer and inoculated with anti-GFP Magnetic Beads, then the accumulation of CaOMP24-myc was detected by Western blot in coprecipitated protein.
The red asterisks indicate the expected band sizes. mutants, CaOMP24 , CaOMP24 , and CaOMP24 , CaOMP24 had a higher infinity to CaOMP24  than the other two (Fig. S16B, see online supplementary material).
These results show that the key region of CaOMP24 for selfinteraction is located at the N-terminal 1-80aa, and especially the N-terminal 1-28aa.
Because the N-terminal 1-28 aa of CaOMP24 is important both for self-interaction and for that with PMMoV CP, we analysed whether PMMoV CP interfered with the self-interaction of CaOMP24 by competitively binding the N-terminal 1-28 aa of CaOMP24. In competitive Co-IP experiments, PMMoV CP-Flag was co-infiltrated with the combination of CaOMP24-GFP and CaOMP24-myc. pGUS-Flag was used as the control. The CaOMP24myc immunocaptured by GFP magnetic beads was significantly reduced by expression of PMMoV CP-Flag (Fig. 8F). Consistently, in the LCI experiment, the signal generated by CaOMP24 selfinteraction was weakened in the presence of PMMoV CP (Fig. S17, see online supplementary material). These results revealed that PMMoV CP impaired CaOMP24-induced PCC by interfering with the self-interaction of CaOMP24.

Discussion
Numerous studies have demonstrated the important role of chloroplasts in plant-virus interactions [10,60]. Many chloroplastrelated proteins have also been identified to be involved in viral replication, movement, pathogenicity, and plant antiviral immunity. Although chloroplast membranes can be targeted and rearranged by several viruses to serve as viral replication sites [12,35], the function of proteins located at the chloroplast membrane in the chloroplast-virus interactions is not well studied. Here, we identified OMP24 that played defense roles against viruses by inducing stromules, PCC, and ROS accumulation, which broadens our knowledge of the function of OMPs in plants.
Chloroplasts possess a double membrane system, consisting of outer and inner envelope membranes. Proteins targeted to the outer envelope membrane are classified into several groups, signal-anchored (SA), tail-anchored (TA), β-barrel, and other proteins [19]. SA and TA proteins contain a single α-helix transmembrane domain located at their N-and C-terminals, respectively [19,31,37]. The moderately hydrophobic TMD is important for SA and TA proteins targeting to the outer membrane [32]. OMP24 has a single transmembrane domain ( Fig. 4; Figs S7 and S8, see online supplementary material) but, unlike SA and TA proteins, this TMD is located in the middle of the sequence towards the C-terminal (Figs S7 and S8, see online supplementary material). Similar to SA and TA proteins, the hydrophobic TMD is also essential for OMP24 targeting to the outer membrane ( Fig. 5; Fig. S9, see online supplementary material). The TMD of OMP24 protein is highly conserved in C. annuum, N. benthamiana, A. thaliana, and S. oleracea (Fig. S1, see online supplementary material), indicating that the TMD may play important roles in the functions of plant OMP24 proteins. Besides the TMD, SA and TA proteins need other structural features to anchor in the membrane, including a Cterminal positively charged region (CPR) in SA and a C-terminal sequence (CTS) in TA [40,46]. In contrast, it seems as if the TMD of OMP24 localizes at the membrane by itself, without the assistance of other segments ( Fig. 4; Fig. S8, see online supplementary material). These results suggest that OMP24 may differ from SA and TA proteins.
The correlation between perinuclear clustering of plastids and immune responses is conserved in animals and plants. In mammalian cells, perinuclear mitochondrial clustering results in ROS accumulation in the nucleus and is important for hypoxiainduced transcriptional regulation and heat shock response [2,3]. In addition, perinuclear mitochondrial clustering has a role in the regulation of cellular calcium transport [47]. Interestingly, overexpression of mitochondrial outer-membrane proteins, such as EXD2 and hFis1, results in mitochondrial perinuclear clustering  [20,30]. In plants, PCC is a general response of plants under biotic and abiotic stresses [16,17,42]. In N. benthamiana, expression of the Xanthomonas effector XANTHOMONAS OUTER PROTEIN Q (XopQ) or TMV p50 activates ETI immunity accompanied by chloroplast relocation around the nucleus [49]. Stromules are associated with PCC and are thought to guide chloroplasts to the nucleus and to transfer the retrograde signaling such as ROS from chloroplasts to the nucleus to regulate resistance gene expression [13,16,26,27]. In our recent work, we identified that upregulated NbNdhM contributed to plant defense against TuMV by inducing PCC and stromules [58]. We here provide evidence demonstrating that OMP24 induces stromules, PCC, and accumulated ROS, and contributes to plant defense. Both results further support the conclusion that the PCC-associated retrograde signaling pathway plays an essential role in plant defense against viruses.
Viruses have numerous strategies to regulate or evade plant immune responses for their own successful infection [11,54]. Emerging evidence suggests that the chloroplast is a central player in perception of pathogen invasion and innate immunity by functioning as a source of defense signals, including phytohormones, ROS, and calcium (Ca 2+ ), and that many pathogens have strategies to attenuate chloroplast-mediated immunity [10,36,42,60]. For instance, re-localization of tomato yellow leaf curl virus (TYLCV) C4 protein from the plasma membrane to the chloroplast aids viral infection as the C4 protein can suppress chloroplast-specific defense mechanisms, specifically the biosynthesis of salicylic acid [45]. TuMV Vpg impairs the antiviral PCC induced by NbNdhM [58]. TuMV P1 interacts with chloroplast protein cpSRP54 to suppress JA-mediated resistance [33]. Barley stripe mosaic virus γ b protein disrupts chloroplast antioxidant defenses by interfering with interactions between NTRC and 2-Cys Prx [53]. We here showed that PMMoV CP suppresses the antiviral defense mediated by OMP24 by interfering with the selfinteraction of OMP24 (Fig. 9), which provides evidence that the viral structural protein CP also plays anti-defense roles by this mechanism. Meanwhile, CPs of TuMV and PVX could not interact with OMP24 (Fig. S18, see online supplementary material). It is possible that other viral factors of both viruses might play antidefense roles.

Plant growth conditions and virus inoculation
N. benthamiana and C. annuum L. cultivar Haonong11 plants were grown in growth chambers with 14-h-light/10-h-dark photoperiod at 24-26 • C.
Infectivity assays in N. benthamiana were performed using PMMoV-GFP, PMMoV, TuMV-GFP, or PVX-GFP infectious clones. Virus symptoms were monitored daily and the GFP f luorescence resulting from viral infection was observed under UV light. For PMMoV inoculation in pepper, sap of N. benthamiana leaves infected with PMMoV was inoculated onto leaves of pepper seedlings by mechanical inoculation.

Plasmid construction
The full-length CDs of OMP24 were amplified by PCR using KOD FX DNA Polymerase (Toyobo, Osaka, Japen) from cDNA of N. benthamiana and C. annuum. The fragments of NbOMP24 TMD , NbOMP24 F109K , CaOMP24 TMD , and CaOMP24 F118K were obtained by Overlapping PCR.
For the yeast two hybrid (Y2H) assays, PMMoV CP, NbOMP24, CaOMP24 and its mutants were inserted into the bait vector (pDHB1) or the prey vector (pPR3-N) using the SfiI restriction site, respectively.
A partial fragment of NbOMP24 (216 nt) and CaOMP24 (283 nt) was amplified and cloned into pTRV2, producing the vectors pTRV2:NbOMP24 and pTRV2:CaOMP24, respectively. To generate hairpin silencing construct targeting NbOMP24, the sequence of NbOMP24 used for VIGS was inserted into pFGC5941 in both sense and antisense orientations to construct the hpNbOMP24 vector. The hpGUS vector used for the control was as previously described [33].
All primers used for plasmid construction are listed in Table S1 (see online supplementary material).

Virus-induced gene silencing in C. annuum and N. benthamiana
To silence NbOMP24, pTRV2:NbOMP24 was co-expressed with pTRV1 in N. benthamiana plants by agroinfiltration. Empty vector (pTRV2:00) combined with pTRV1 served as the control. Plants silenced for 12 days were inoculated with PMMoV-GFP. For silencing of CaOMP4 in pepper, pTRV2:CaOMP24 and pTRV1 were mixed at a 1:1 ratio and agro-infiltrated onto 2-leaf stage pepper leaves. Pepper plants silenced for 18 days were used for further analysis.

Split ubiquitin yeast two-hybrid assays
The yeast two-hybrid assays based on the split-ubiquitin technique were performed using the DUAL hunter starter kit (Clonetech, Mountain view, California, USA). PMMoV CP, NbOMP24, CaOMP24 and its mutants were inserted into the pDHB1 or pPR3-N vectors, followed by transformation into the NMY51 yeast strains by a lithium acetate method. The transformed yeast cells were then plated on SD/−Leu/−Trp to verify successful co-transformation and then transferred onto SD/−Leu/−Trp/-His/−Ade dropout medium to assess the interaction of the target proteins.

Western blot and co-immunoprecipitation (co-IP) assays
For western blot, total proteins were extracted from pepper or N. benthamiana leaves with lysis buffer. The proteins were separated by SDS-PAGE, and subsequently transferred onto Immobilon ® -P PVDF Membrane (Merck, Co. Cork, Ireland) using the wet transfer method. The membranes were probed with primary antibody (Myc, GFP, Flag, PMMoV CP, TuMV CP, and PVX CP) and HRP-conjugated (anti-Mouse or anti-Rabbit) secondary antibodies (TransGen Biotech, Beijing, China) followed by ECL detection using Immobilon Western HRP Substrate (Millipore, Darmstadt, Germany). The Rubisco large subunit protein (RbcL) stained with Coomassie Brilliant Blue R-250 Dye was used as a loading control.
The co-IP assays were performed as described previously [34]. The proteins to be tested were transiently expressed by Agrobacterium infiltration in N. benthamiana plants. Total proteins from infiltrated N. benthamiana leaf tissue were extracted and incubated with GFP-Trap ® beads (ChromoTek, Planegg-Martinsried, Germany). The precipitates were washed at least three times with GTEN buffer containing 0.1% NP-40 and analysed by Western blotting using anti-GFP, anti-Myc, or anti-Flag antibodies.

qRT-PCR analysis
Total RNA was isolated from pepper or N. benthamiana leaves using Trizol reagent (Invitrogen, Carlsbad, California, US). The firststrand cDNA was synthesized using the ReverTra Ace™ qPCR RT Master Mix with gDNA Remover kit (Toyobo, Osaka, Japen). qRT-PCR was conducted on the LightCycler ® 480 Real-Time PCR System (Roche, Mannheim, Germany) using SYBR Green qPCR Master Mix (Vazyme, Nanjing, China) in accordance with the manufacturer's instructions. The results were analysed by the CT method [43]. C. annuum β-Tublin or Nbactin was used as the reference gene. At least three biological replicate samples were used. The primers used for qRT-PCR are listed in Table S1 (see online supplementary material).

Laser confocal microscopy assays
Subcellular localization and bimolecular f luorescence complementation (BiFC) assays were performed as described previously [58]. The agro-infiltrated N. benthamiana leaves were observed at 24 to 48 hpi using a Nikon A1 confocal microscope.

Generation and identification of NbOMP24 transgenic N. benthamiana plants
The NbOMP24 transgenic N. benthamiana plants were generated by leaf disk transformation with Agrobacterium containing pCV-NbOMP24.1-Myc [44]. Western blot with anti-myc antibody (TransGen Biotech, Beijing, China) was performed to screen the transgenic plants.

Detection of cellular ROS by DCFH-DA staining
Reactive oxygen species (ROS) accumulation measured by DCFH-DA staining was performed as described previously [50]. A 100 mM stock solution of DCFH-DA (MCE, Shanghai, China) was prepared by dissolving 100 mg in 2.0522 mL DMSO. Leaf discs from agroinfiltrated and transgenic plants overexpressing NbOMP24 were stained with 50 μM DCFH-DA in 10 mM Tris-HCl (pH 7.5) by vacuum-infiltration and inoculated in the darkness for 30 min. ROS accumulation was visualized as green f luorescence in a Nikon A1 confocal microscope with excitation at 488 nm and emission at 515-530 nm.