Endocytosis-mediated entry of a caterpillar effector into plants is countered by Jasmonate

Insects and pathogens release effectors into plant cells to weaken the host defense or immune response. While the imports of some bacterial and fungal effectors into plants have been previously characterized, the mechanisms of how caterpillar effectors enter plant cells remain a mystery. Using live cell imaging and real-time protein tracking, we show that HARP1, an effector from the oral secretions of cotton bollworm (Helicoverpa armigera), enters plant cells via protein-mediated endocytosis. The entry of HARP1 into a plant cell depends on its interaction with vesicle trafficking components including CTL1, PATL2, and TET8. The plant defense hormone jasmonate (JA) restricts HARP1 import by inhibiting endocytosis and HARP1 loading into endosomes. Combined with the previous report that HARP1 inhibits JA signaling output in host plants, it unveils that the effector and JA establish a defense and counter-defense loop reflecting the robust arms race between plants and insects.


Introduction
Plants recognize pathogen/insect/damage associated molecular patterns (PAMPs/HAMPs/DAMPs) to trigger defense response in plant-biotic interactions [1][2][3][4][5][6][7] .Jasmonate (JA) is one of the key regulators in plant defense against insects and pathogens 8-11 .On the other hand, insects and pathogens secrete effectors to suppress plant defense for tness 12,13 .The rst reported insect effector is the Glucose Oxidase (GOX) identi ed from the oral secretion (OS) of Helicoverpa Zea which inhibits the woundinginduced nicotine accumulation in tobacco 14,15 .However, most of the reported insect effectors are from piercing-sucking insects 16,17 .C002-like proteins is rstly found in Acyrthosiphon pisum saliva and ubiquitously exist in various species of aphids which in uence the aphid feeding behaviors and increase their reproduction on host plants 18,19 .Bt56 is an effector in the white y OS, which promotes insect infection in tobacco 20 .Besides proteins, RNA molecules can also act as effectors.The noncoding RNA Ya of Myzus persicae is delivered into plants through aphids' feeding sites and migrated systemically to suppress plant immunity 21 .
Migrations of heterogenous regulators into host plants are essential to perform their functions and are always achieved by diffusion or passing through transporters in the plasma membrane, or by secretion systems like exocytosis and endocytosis tra cking [22][23][24][25] .Many bacterial pathogens directly inject effectors into host cells by specialized secretion machineries 26 .RxLR-like effectors of oomycetes bind to the phospholipid, phosphatidylinositol-3-phosphate (PI3P) and enter host cells through lipid raft-mediated endocytosis 27 .Extracellular vesicles (EVs) are important to transfer substances such as proteins, RNAs, lipids and metabolites across kingdoms [28][29][30] .Patellins (PATLs) and Tetraspanins (TETs) are found in exosomes suggesting their functions in cell-to-cell tra cking 31 .Botrytis cinerea infection induces plants to secrete TET8-related exosomes which carries plant derived sRNAs and triggers cross-kingdom RNAi 32 .CTL1 (choline transporter-like 1) is another component in vesicle tra cking system which is involved in the tra cking of PIN proteins, the auxin e ux transporters 33 .
Aphids deliver their effectors into plants when the stylets stick into cells, however, the nal localization of effectors are quite different.For example, Myzus persicae effector, Mp10 is distributed in the cytoplasm and chloroplast of mesophyll cells while MpPIntO1 and MpC002 are in the sheath-like structure near the feeding sides 34 .These results indicate that the translocation of insect effectors in plants is speci cally regulated.However, the mechanisms of insect effector imports in plants are largely unknown.In our previous work, we identi ed an effector HARP1 from the cotton bollworm (Helicoverpa armigera) OS which suppressed JA response in plants 35 .HARP1-like proteins are widely present in Lepidoptera and their functions are likely conserved in Noctuidae.In this study, we demonstrated that HARP1 entered the plant cells through endocytosis and the vesicle tra cking related proteins CTL1, PATL2 and TET8 were required for its entry.Interestingly, endocytosis-mediated entry of HARP1 into plants was countered by JA.JA restricted the HARP1 import by inhibiting endocytosis and HARP1 loading on endosome, thereby creating a counter-defense response.

HARP1 tends to be granulated and is moving in plants
In our previous work, we found that HARP1 from cotton bollworm was an effector to interfere with plant wounding responses by interacting with JAZs 35 .To trace the HARP1 transportation in plants, we generated Venus-HARP1 (V-HARP1) fusion protein for visualization.Confocal microscopy observation shows that V-HARP1 is found around the leaf wounding sites of Arabidopsis, while Venus alone is barely detected.Furthermore, the transportation is dose associated, higher concentration of V-HARP1 leads to stronger uorescence intensity in leaf cells.At the moderate concentration (0.1 mg/ml), the HARP1 signals could be still detected in cells after incubation though the uorescence is weaker than that of 1 mg/ml (Supplementary Fig. 1a).For better observation, we choose the V-HARP1 with concentration of 1 mg/ml, an acceptable protein concentration also used in the study of fungal effector transportation 27 , for all the subsequent studies.Application of V-HARP1 on the leaf wounding sites impairs the inductions of gene expressions in plant wounding response (Supplementary Fig. 1b), indicating that the effector activity is not affected by N-terminal fusion of Venus.Besides Arabidopsis, V-HARP1 could also get into a variety of plants including cotton, tobacco, and rice (Supplementary Fig. 2).Transmission electron microscopy (TEM) observations show that it was V-HARP1 not Venus alone could be found in the pavement and mesophyll tissues of Arabidopsis (Fig. 1a and Supplementary Fig. 3).Under confocal microscopy, V-HARP1 in plant leaves tended to cluster as granules with an average size of 1.2 µm 2 (Fig. 1b and Supplementary Fig. 4a).Interestingly, about 44% of total V-HARP1 granules were moving (Supplementary Fig. 4b and Supplementary Movie 1).This gives a hint that HARP1 entry into cells might be mediated by vesicle tra cking.

HARP1 locates in endosomes
To exclude the possibility that the granulated character of V-HARP1 in plant cells was caused by Venus, we than observed Venus distribution in plants.Since Venus alone cannot get in plant cells automatically like V-HARP1, we than delivered Venus into leaf tissue by injection and found most of Venus was uniformly distributed.And, when V-HARP1 was injected into plant tissues, the granulated V-HARP1 was observed (Supplementary Fig. 5).Furthermore, using N-(3-Triethylammoniumpropyl)-4-(6-(4-(diethylamino)phenyl)hexatrienyl) pyridinium dibromide (FM4-64), an endocytosis tracer, to indicate internalized endosomes 36,37 , we found that some V-HARP1 granules were colocalized with FM4-64stained endosomes (Supplementary Fig. 5).Similarly, when V-HARP1 were incubated with wounded leaves, colocalization of V-HARP1 granules with FM4-64-stained endosomes were also observed (Supplementary Fig. 6) and part of the V-HARP1-loaded endosomes were moving (Supplementary Movies 2, 3).

HARP1 import is mediated by endocytosis
Endosome recycling inhibitor, brefeldin A (BFA) blocks vesicle tra cking from the TGN to PM leading to the aggregations of TGN 45 and the formation of enlarged endosomes, also called BFA bodies 46 .When the BFA pretreated Arabidopsis leaves were incubated with V-HARP1, the uorescent signals of V-HARP1 were observed around the wounding sites as that observed from the mock samples (Fig. 2a).However, when focused on the endosomes, some V-HARP1-loaded endosomes tended to aggregated after BFA treatment (Fig. 2c, d).
Tyrphostin A23 (A23) and Wortmannin (Wm) are endocytosis inhibitors 47,48 .To con rm whether endocytosis was required for HARP1 import, the wounded Arabidopsis leaves were pretreated with A23 and Wm to shut down the endocytosis process and then incubated with V-HARP1.It was observed that the uorescent signals of V-HARP1 around the wounding sites were obviously reduced by A23 or Wm pretreatments compared to those of the mock treatment (Fig. 2a, b).Furthermore, though the V-HARP1 could still merge with the FM4-64 labeled PM, the punctate patterns of V-HARP1 were substantially declined and the granulated V-HARP1 was hardly detected inside the cells by the A23 or Wm pretreatments (Fig. 2c, d).These results suggested that HARP1 import required endocytosis.Import of V-HARP1 is reduced in some vesicle tra cking mutants CTL1, Sphingolipids Delta-8 desaturase (SLD), PATLs and TETs regulate endomembrane systems and are involved in maintaining membrane lipid homeostasis and vesicle tra cking 31-33,49−55 .We then analyzed whether HARP1 import was affected in these vesicle tra cking-related mutants and found that the V-HARP1 import was signi cantly reduced in ctl1, sld1 sld2, patl2 and tet8 while barely affected in 35S::PDLP5, patl1 and patl3 (Fig. 3a, b and Supplementary Fig. 8).Regarding that PDLP5 (Plasmodesmata-located protein 5) negatively regulates the permeability of plasmodesmata by enhancing callose deposition and overexpression of PDLP5 in plants (35S::PDLP5) reduced the permeability of plasmodesma 56 , these results excluded the possibility that the V-HARP1 import were mainly from the plasmodesmata.Successful import into plant cells is prerequisite for HARP1 to meet JAZ proteins and subsequently suppress the inductions of the JA response genes.We tested the V-HARP1 effects on the wounding response in ctl1, patl2 and tet8.In general, in these mutants, wounding could induce the expressions of JA response genes but the inductions were reduced to some extent.And with the V-HARP1 applications, the inductions were signi cantly reduced in wild type while hardly affected in ctl1, patl2 and tet8 (Fig. 3ce).Moreover, the expressions of CTL1, PATL2 and SLD2 were induced by application of V-HARP1 to the leaf wounding sites when compared with the Venus application (Supplementary Fig. 9).Feeding test showed that larvae growth was retarded on ctl1 and patl2 than that on wild type while not obviously effected on tet8 and better on 35S:PDLP5 (Supplementary Fig. 10).These results indicated that vesicle tra cking mediated by SLD1/SLD2, CTL1, PATL2 and TET8 were involved in V-HARP1 import.
HARP1 targets to CTL1, PATL2 and TET8 for successful import CTL1 and TET8 are transmembrane proteins while PATL2 are membrane-associated proteins, furthermore, TET8 and PATL2 were found in exosome [30][31][32] .Taken together, these proteins have the possibility to meet HARP1.We then analyzed whether CTL1, PATL2 and TET8 could directly interact with HARP1.As transmembrane proteins, the outside membrane domains of CTL1 and TET8 are most likely to bind to HARP1.The predicted structure of CTL1 contains ve extracellular loops (ECs), of which EC1 is the longest loop with 182 amino acids 57 (Supplementary Fig. 11a) and TET8 contains a very short EC1 loop and a longer EC2 loops with 139 amino acids 30 (Supplementary Fig. 11b).PATL2 has a C-terminal GOLD domain (PATL2C110) which is predicted to mediate protein-protein interactions 51 .The yeast two hybrid (Y2H) assay revealed that HARP1 could interact with CTL1EC1, PATL2C110 and TET8EC2, respectively (Supplementary Fig. 12a).And in pull-down assay, CTL1EC1, PATL2C110 and TET8EC2 coimmunoprecipitated with HARP1 (Fig. 4a).Moreover, full-length of CTL1, PATL2 and TET8 interacted with HARP1 in Bilayer Luciferase Complementation (BiLC) assay (Supplementary Fig. 12b).
To know which domain of HARP1 is required for its interaction with CTL1, PATL2 and TET8.We tested the binding activity of the truncated HARP1 (Supplementary Fig. 14a) with CTL1EC1, PATL2C110 and TET8EC2.In Y2H assay, it revealed that the deletion of the ve amino acids of the HARP1 C-terminal (V-HARP1δC5) and the deletion of 44 or more amino acids of the HARP1 N-terminal (V-HARP1δN44 and V-HARP1δN49) abolished its interaction with the CTL1EC1, PATL2C110 and TET8EC2, while deletion of 39 or less amino acids of the HARP1 N-terminal (V-HARP1δN34 and V-HARP1δN39) had no obvious effects on the protein interactions (Supplementary Fig. 14b-d).Consistently, V-HARP1δN39 could be still coimmunoprecipitated with CTL1EC1, PATL2C110 and TET8EC2, while V-HARP1δN44 could not in pulldown assay (Fig. 4b).We then detected the imports of truncated HARP1 into plant cells.Interestingly, only V-HARP1δN34 and V-HARP1δN39 but not V-HARP1δC5, V-HARP1δN44 and V-HARP1δN49 could be successfully imported (Fig. 4f and Supplementary Fig. 15).These results suggested that both the amino acid residues 40-49 and 117-122 were required for the interactions of HARP1 with CTL1, PATL2 and TET8 proteins.We then tested the effector activity of truncated V-HARP1 on plant wounding response and found that host defense responses were suppressed by V-HARP1δN34 and V-HARP1δN39 but not V-HARP1δN44 treatments (Fig. 4g).This indicated that the interactions of HARP1 with CTL1, PATL2 and TET8 proteins were required for its successful import into plant cells to perform its function.

Plant defense hormone JA counters HARP1 import
It is well known that immune response is required to get rid of the invaded foreign proteins in animals 58 .We wonder whether the HARP1 import into plants would be restricted by host immunity.As the main defense hormone, JA and JA-Ile shoot up in plant response to insect herbivory or mechanical wounding 59-61 and the wounding response in the JA synthesis de cient mutant, aos, are largely reduced 62,63 .From our transcriptome data of wild type and aos, about 92 genes annotated as response to JA were detected (Supplementary Table 2) and consistently, their total expressions were higher in wild type than those in aos two hours post wounding.Among these genes, 37 were signi cantly higher in wild type while only three were the opposite (Supplementary Fig. 16a, b).However, some other genes (1121) exhibited higher expressions in aos than those in wild type in response to wounding (Supplementary Fig. 16c) and were signi cantly enriched in GO items of secretory vesicle, membrane, cell wall, extracellular region, and endoplasmic reticulum (Fig. 5a).The ClueGO network showed that cell wall, extracellular region and endoplasmic reticulum were closely associated with membrane related items (Fig. 5b).Consistently to the RNA-seq analysis, qRT-PCR revealed the expressions of four selected genes (EXLA2, GRP-5, EARLI1, PKS2) encoding membrane related proteins displayed higher induction in aos upon wounding than those in wild type (Fig. 5c).Together, these results suggested that JA might negatively associate with membrane related functions.As membrane proteins are important in endocytosis and vesicle homeostasis 64-66 , we further analyzed whether JA had impacts on endocytosis.Normally, the internalized endosomes traced by FM4-64 were less in the wounded leaves of wild type than those of aos.With the presence of MeJA, the accumulation of internalized endosomes was signi cantly reduced to the similar degree between wild type and aos (Fig. 5d).These results indicated that JA negatively regulated endocytosis.
We then investigated whether HARP1 import was affected by JA.After incubation with the wounded leaves, the uorescence signal of V-HARP1 was weaker in JA hypersensitive mutant jazQ, stronger in aos and similar in the JA insensitive mutant coi1-2 when compared with that in the wild type (Fig. 6a and Supplementary Fig. 17).With the MeJA pretreatment, the V-HARP1 signal was signi cantly reduced in wild type and aos, while not obviously affected in coi1-2.For jazQ, the V-HARP1 signal was constitutively low with or without MeJA pretreatment (Fig. 6a and Supplementary Fig. 17).These results suggest that JA signaling negatively regulates HARP1 import.
When focusing on endosomes, we found that the counts of both total and V-HARP1-loaded endosomes were higher in aos and coi1-2 than those in wild type by wounding treatment.Upon wounding and MeJA treatment, the total and V-HARP1-loaded endosomes dramatically reduced in wild type and aos while no signi cant change was found in coi1-2.For jazQ, the endosomes were extremely low in both consequences (Fig. 6b-d).Interestingly, we found the proportion of V-HARP1-loaded to the total endosomes signi cantly dropped in wild type and aos upon MeJA pretreatment compared to those without pretreatment (Fig. 6e).This indicated that besides inhibition of endocytosis, JA also affected HARP1 loading on endosomes.

Discussion
Although plants have sophisticated cuticular barriers, insect effectors can get into plant cells and this is often required for their functions.Compared with the studies on the mechanism of how effectors inhibit plant immune or defense response, the studies on the imports of effectors into plant cells are very limited.Here, we found that H. armigera effector HARP1 got into plant cells via endocytosis and its binding activities with CTL1, PATL2 and TET8 were required for its tra cking.HARP1 was likely internalized into plant cells by interacting with CTL1, PATL2 and TET8.Since PATL2 and TET8 were also found in extracellular vesicles (EV) and involved in cell-to-cell tra cking 31,32,52 , it is possible for HARP1 to achieve intercellular transport.After entering plant cells, HARP1 further interacted with JAZ to block JA response in defense against insect.On the other hand, JA restricted HARP1 import by inhibiting endocytosis and HARP1 loading on endosomes to establish a counter-defense response (Fig. 7) and this revealed a new role of JA in the arms race between plants and insects.In animals, the invade proteins are monitored and removed by immune system.Here in plants, the entry of insect effectors is also restricted by JA signaling and this re ects the conserved way of plants and animals in immune response.
In plant, vesicle tra cking was involved in transport of toxic substances to resist pests and pathogens 67,68 .Here, we found that the insect hijacked the plant vesicle tra cking system to deliver their own effectors.It seems that both insects and microbial pathogens use the similar strategies to deliver their effectors to host plants re ecting the evolutionary similarity.The RxLR-like effectors of lamentous pathogens, such as oomycetes and fungi, enter plant cells by lipid raft mediated endocytosis 27 , and the conserved motifs like RXLR, YKARK, and RGD are required for their successful entry 69 .It is speculated that the import of this type effector could also be mediated by certain membrane-associated proteins or transporters though it has not yet been identi ed 70 .Unlike these microbe pathogen effectors, there was no single motif that was responsible for HARP1 transportation.Never the less, both the residues 40-49 and 117-122 are required for successful HARP1 import (Fig. 4f) and this implies differences between transportation of microbe pathogen and insect effectors in plants.
Introducing exogenous molecules into plant cells is often required in plant research.The most used methods are Agrobacterium transformations and gene guns.Exogenous molecules can also be introduced into plant cells with the help of cell-penetrating peptides and nanoparticles 71,72 .In another case, Ustilago maydis, the fungal pathogen was developed to deliver the bioactive host protein to alter maize anther cell behavior in Situ 73 .In our study, HARP1-fused Venus can be delivered into cells of various plant species.This suggests that it is possible to develop the HARP1-based delivery system for exogenous molecules import into plants.
The seeds of Arabidopsis thaliana, tobacco plants (Nicotiana benthamiana) and cotton plants (Gossypium hirsutum) used in this study were grow in long days (16-h light/8-h dark) under 22℃.
The rice (Oryza sativa) seeds were washed with 75% ethanol for 1 minute, and with 20% (v/v) NaClO for 30 minutes, then rinsed ve times with sterile water.The mature embryos were grown on N 6 D 2 medium and cultured in dark at 28℃ for 14 days to induce callus.

Plant Treatment
HARP1 was identi ed from the oral secretions of cotton bollworm (Helicoverpa armigera) 35 and the sequence information can be found in NCBI (https://www.ncbi.nlm.nih.gov/protein/XP_047035071.1?report=genbank&log$=protalign&blast_rank=1&RID=F3YF3FK8016).To detect the effector activity of V-HARP1, the second pair of true leaves at the rapid expanding stage (about 18-day-old seedlings) were wounded (punched 3 ~ 4 holes per leaf) and the unwounded leaves were used as negative control, the V-HARP1 and Venus protein solutions (1 mg/ml) were painted to the wounding sites immediately.Four hours later, leaves were harvested and used for qRT-PCR analyses of the indicated wounding induced gene expressions.
To observe the import of HARP1 in plants, the leaves of Arabidopsis, cotton and tobacco were cut into small pieces with a hole in the center and incubated with V-HARP1 or its truncated forms (1 mg/ml) for 2-4 hours.Samples were then washed with PBS containing 0.08% BSA for at least 5 times, 20 minutes for each time and detected under confocal microscopy.As a negtive control, the Venus (1 mg/ml) were used for incubation with plant leaves like V-HARP1 and followed by the same washing procedure to insure get rid of all the extra proteins adhering to the surface and subsequently for confocal microscopy.Under our experimental condition, Venus along cannot be detected in leaf cells.
For the assay of jasmonate effects on V-HARP import, MeJA (Sigma-Aldrich) was dissolved in ethanol and diluted in double-distilled water to a nal concentration of 50 µM.As the mock treatment, water solutions with equal volumes of ethanol were used.18-day-old Arabidopsis were sprayed with water solutions of MeJA and ethanol, respectively.Two hours later, detached leaves were punctured and incubated with the V-HARP1 solutions for 2-4 hours and then observed under confocal microscopy.

Confocal Microscopy
Venus was fused to the N-terminal of HARP1 (V-HARP1) for visualization.After incubation with Venus or V-HARP1 followed by 4-5 times washing, samples were detected by confocal microscopy.To detect HARP1 in cells, regions 100 ~ 600 µm from the wounding site were selected for observation (Supplementary Fig. 18).We found that the uorescent signal from the Venus treated samples were hardly to detect, therefore, excluding auto uorescing which might be caused by wounding and the possibility that Venus itself could enter plant cells under our experimental condition.To label plasma membrane and trace the internalized endosomes, samples were incubated with 2 µM FM4-64 (Invitrogen) in the dark for 5-7 minutes, and washed twice with double-distilled water.
All images and movies were taken by confocal systems, Leica SP8 or Olympus spinSR equipped with different immersive objectives (Leica SP8 with 20XW/NA 0.75 and 60XW/NA 1.2, Olympus spinSR with 30XSil/NA 1.05 and 60XSil/NA 1.30).The confocal z-axis resolution range is about 1240 ~ 647 nm.The moving HARP1 granules were traced by time series at a xed layer.For multi-channel images, we use sequential imaging procedure to exclude uorescence signals crosstalk.
The excitation/emission wave lengths for YFP and FM4-64/mRFP/mCherry signals were 515/525-560 and 561/610-650 nm, respectively.And Images were captured with strictly identical acquisition parameters for the quantitative uorescence intensity.The uorescence signal intensity was calculated using ImageJ software and statistically analyzed with GraphPad Prism software.

Immuno-localization
After the Arabidopsis leaves were incubated with Venus or V-HARP1 followed by 4-5 times washing, samples were cut into pieces (1-2 mm 2 ) and immediately xed in 4% (w/v) paraformaldehyde in phosphate-buffered saline (PBS) at 4℃ overnight.Fixed leaves were embedded in LR White resin (Sigma-Aldrich) after dehydration through a graded alcohol series.Sections were prepared on a Leica Microsystem UC7 ultramicrotome.Ultrathin sections of approximately 70 nm were mounted onto nickel grids for immuno-gold labeling.
Grids were observed under 120 kV transmission electron microscope (TEM) after staining.The replicate sections of Venus treated samples had no signals of immuno-gold in cells indicating that the signal is speci c and Venus alone could not enter plant cells.
The constructed vectors were transformed into Escherichia coli strain BL21 (DE3) strain.0.25 mM isopropyl β-D-thiogalactopyranoside (IPTG) was used to induce protein expression.E. coli cells carrying the designed vectors were cultured in 37℃.After IPTG addition, the cells were transferred to 16℃ for another 12-16 hours before harvested.The His fusion proteins were puri ed by Ni a nity column (Ni-NTA resin, Qiagen) and the GST fusion proteins were puri ed by Glutathione Sepharose 4B resin (GE Healthcare).The eluted proteins were then concentrated and desalted using an Amicon Ultra-15 Centrifugal Filter Unit (10,000 molecular weight cutoff [MWCO], EMD Millipore) with 50 mM Tris-HCl (pH = 8.0) to a nal protein concentration of 1 mg/ml.Transient expression assay in Nicotiana benthamiana (N.benthamiana) About 4-5 monoclones of Agrobacterium carrying the target genes were grown together overnight in liquid LB medium under 30 ℃, then were transferred to the new liquid LB medium at a ratio of 1:500 and grown overnight.The cultures were centrifuged at room temperature and cells were resuspended with in ltration buffer (10 mM MgCl 2 , 10 mM MES, 150 µM Acetosyringone, pH = 5.7) at an OD 600 of about 1.0.The cell solutions were in ltrated into the back of tobacco leaves.2-3 days later, leaves were harvested for BiLC and pull-down assay.

Bimolecular Luciferase Complementary (BiLC) Assay
The BiLC assays were performed as described 78 .For HARP1 interaction with CTL1 and PATL2, CTL1 and PATL2 were fused to the amino-terminal half of LUCIFERASE (nLUC), HARP1 was fused to the carboxylterminal half of LUC (cLUC).For HARP1 interaction with TET8, HARP1 was fused to nLUC and TET8 was fused to cLUC.cLUC and nLUC alone were used as controls.The oligonucleotide primers used for these vectors are given in Supplementary Table 1.Agrobacterium cell solutions were in ltrated into tobacco leaves.Luciferin (1 mM) was in ltrated before LUC activity was monitored after 2 d.

Yeast Two Hybrid
V-HARP1 variants were introduced into the pGBKT7 (Clontech).CTL1EC1, PATL2C110 and TET8EC2 were introduced into the pGADT7 (Clontech).The oligonucleotide primers used for these vectors are given in Supplementary Table 1.A LiCl polyethylene glycol method was used to transfer the indicated plasmids into yeast strain AH109 (Clontech).Transformants were grown on SD-Leu-Trp mediums for 2-3 days and then tested on SD-Leu-Trp-His mediums (-L-T-H) or SD-Ade-Leu-Trp-His mediums (-A-L-T-H) with the indicated 3-amino-1,2,4 triazole (3-AT).At least 10 individual clones for each transformant were analyzed to con rm the interactions.

Insect Feeding Test
The cotton bollworm (Helicoverpa armigera) larvae were obtained from the Institute of Zoology, Chinese Academy of Science.About 30 synchronous third instar larvae were fed on 60 Arabidopsis plants of 20 days old.After fed on for 3-4 days, weight increases were recorded.

RNA-seq And Transcriptome Analysis
The leaves of 10-day-old wild type (Col-0) and JA synthesis mutant (aos) were wounded (W) and harvested 2 hours post wounding.The untreated plant leaves (CK) were used as control.Total RNA was extracted using the Plant RNA puri cation kit (Qiagen, 74904).Library construction and RNA-sequencing with three biological replicates was performed on an Illumina HiSeqXten platform (Illumina, San Diego, CA, United States) at Majorbio (Shanghai, China).The clean reads were mapped to the Arabidopsis genome (TAIR10) using Hisat2 v2.0.4.HTSeq v0.9.1 to count the reads numbers mapped to each gene.Fragments Per Kilobase of exon model per Million mapped fragments (FPKM) was calculated based on the gene length and the mapped-reads counts.Genes with a fold-change greater than 2 and FDR less than 0.05 were considered as differentially expressed genes (DEGs) using the R package DESeq2 79 .Gene Ontology (GO) enrichment and network analysis were performed by using the Cytoscape v3.8.2 plug-in ClueGO 80 .Signi cant enrichment was de ned as FDR less than 0.05.GO clusters were inferred by Kappa score.The complete list can be found in Supplementary Table 2.The raw data are deposited in the NCBI (BioProject accession number: PRJNA760932) and will be released as soon as publication.The reviewer link: https://dataview.ncbi.nlm.nih.gov/object/PRJNA760932?reviewer=kbk81o7uvur0kpn4dh4kpus2lr.

Gene Expression Analyses
Plant total RNA was extracted by Trizol reagent (Invitrogen).1.5 ug of total RNAs were treated with DNase I (1 unit per ul; Fermentas) and used for cDNA synthesis with oligo (dT) primer (TransGen Biotech).qRT-PCR was performed using SYBR green PCR master mix (TaKaRa) on a real-time PCR system (CFX thermocycler; Bio-Rad, Hercules, CA).S18 in Arabidopsis (At4g09800) was used as an internal standard.
The gene average expression levels were calculated from 2 −ΔΔCt values.At least three biological triplicates with technical triplicates were performed.The oligonucleotide primers for all the genes tested are given in Supplementary Table 1.

Statistical analysis
Data are presented as means ± SEM.Signi cances were examined by Student's t test or by one-way or two-way ANOVA followed by multiple comparison tests with GraphPad Prism software.No less than three independent experiments were performed for each assay, and every experiment contains at least three biological replicates.endosomes and Merge/FM4-64 stands for the ratio of V-HARP1-loaded to total endosomes (e).

Declarations Figures
Figure 7

Figure 2 The
Figure 2

Figure 3 The
Figure 3

Figure 4 The
Figure 4