A novel strategy for improving watermelon resistance to cucumber green mottle mosaic virus by exogenous boron application

Abstract The molecular mode controlling cucumber green mottle mosaic virus (CGMMV)‐induced watermelon blood flesh disease (WBFD) is largely unknown. In this study, we have found that application of exogenous boron suppressed CGMMV infection in watermelon fruit and alleviated WBFD symptoms. Our transcriptome analysis showed that the most up‐regulated differentially expressed genes (DEGs) were associated with polyamine and auxin biosynthesis, abscisic acid catabolism, defence‐related pathways, cell wall modification, and energy and secondary metabolism, while the down‐regulated DEGs were mostly involved in ethylene biosynthesis, cell wall catabolism, and plasma membrane functions. Our virus‐induced gene silencing results showed that silencing of SPDS expression in watermelon resulted in a higher putrescine content and an inhibited CGMMV infection correlating with no WBFD symptoms. SBT and TUBB1 were also required for CGMMV infection. In contrast, silencing of XTH23 and PE/PEI7 (low‐level lignin, cellulose and pectin) and ATPS1 (low‐level glutathione) promoted CGMMV accumulation. Furthermore, RAP2‐3, MYB6, WRKY12, H2A, and DnaJ11 are likely to participate in host antiviral resistance. In addition, a higher (spermidine + spermine):putrescine ratio, malondialdehyde content, and lactic acid content were responsible for fruit decay and acidification. Our results provide new knowledge on the roles of boron in watermelon resistance to CGMMV‐induced WBFD. This new knowledge can be used to design better control methods for CGMMV in the field and to breed CGMMV resistant watermelon and other cucurbit crops.


| INTRODUC TI ON
Watermelon (Citrullus lanatus) is a popular and economically important crop worldwide (Guo et al., 2019). China is currently the largest watermelon-growing country, accounting for about 61% of the global watermelon production in 2019 (http://faost at.fao.org/). Watermelon production in many countries is threatened by various abiotic and biotic stresses, including many plant-infecting viruses.
CGMMV was first reported in cucumber in the UK followed by more than 20 countries in Europe, Asia, and South America (Dombrovsky et al., 2017). In 2007, the Agricultural Ministry of China declared CGMMV as a plant quarantine pest that could potentially threat the production of cucurbit crops in China (http:// www.moa.gov.cn/). In recent years, several research laboratories have analysed the effects of CGMMV infection on gene expression in watermelon plants (Li et al., , 2020Sun et al., 2019).
However, how CGMMV infection affects watermelon fruit growth and development remained largely unknown, due mainly to the long watermelon growth period and lack of an efficient assay method.
Boron is an essential micronutrient for plant growth and development (Brdar-Jokanovi, 2020;Camacho-Cristóbal et al., 2011). The amount of boron in soil is relatively high and the difference between boron deficiency and toxicity is very small.
Currently, several factors, including soil pH, texture, and climatic condition, are known to affect boron uptake by plants (Brdar-Jokanovi, 2020;Shorrocks, 1997). Many reports have shown that boron deficiency can alter various plant physiological processes (Brdar-Jokanovi, 2020;Camacho-Cristóbal et al., 2011;Eggert & Von Wirén, 2017), alter polyamine biosynthesis (Camacho-Cristóbal et al., 2005), and decrease the expression of several cell wall modification-associated enzymes while increasing the production of actin and tubulin in Arabidopsis roots (Camacho-Cristóbal, Herrera-Rodríguez, et al., 2008;Yu et al., 2001). To minimize the damage caused by inadequate boron, plants respond with the synthesis of various hormones (Eggert & Von Wirén, 2017) and activation of signal transduction, stress responses (Lu et al., 2015), and the glycolytic pathway (Zhou et al., 2014). Boron deficiency can be prevented through exogenous boron application to improve crop yield and quality as well as to inhibit plant diseases (Gupta et al., 2017).  (Ananthakrishnan et al., 1989;Shimomura, 1982). Moreover, the grapevine Pinot Gris virus (GPGV)-induced disease symptoms are similar to that induced by boron deficiency, and the expressions of boron homeostasis-associated genes were affected by both GPGV infection and boron deficiency (Buoso et al., 2020).
Because watermelon is recalcitrant to genetic transformation, analyses of gene functions in this plant are hindered. To circumvent this obstacle, we decided to investigate the molecular mechanisms controlling CGMMV infection in watermelon fruits through reverse genetics using a CGMMV-based virus-induced gene silencing (VIGS) pV190 vector as described previously (Lange et al., 2013;Liu et al., 2020). Currently, two VIGS vectors (i.e., the CGMMV-based pV190 vector and a tobacco rattle virus [TRV]based vector) have been reported to be effective in gene silencing in cucurbits (Fang et al., 2021;Liu et al., 2020). In this study, we found that the pV190 vector-induced gene silencing in watermelon fruits lasted more than 54 days and thus is suitable for studies of H2A, XTH23, PE/PEI7, DnaJ11, and ATPS1 were likely to participate in watermelon resistance to CGMMV infection. The contents of many metabolites, including polyamines, celluloses, pectins, malondialdehyde (MDA), and lactic acid (LA), were found to be altered in the watermelon fruits silenced for specific DEGs, and these alterations were found to affect watermelon resistance to CGMMV infection. We consider that the findings presented in this paper provide not only new insights into the role of boron in watermelon resistance to CGMMV but also new knowledge for the establishment of an efficient and environmentally friendly control method for WBFD in watermelon fruits and possibly other cucurbit crops.
F I G U R E 1 CGMMV-induced symptoms in watermelon fruits and virus accumulation. (a) Flowchart of CGMMV inoculation, water or boron application, and sampling. (b) Watermelon blood flesh disease symptoms in watermelon fruits from the phosphate-buffered saline (PBS, P) + water (H)-, P + boron (B)-, CGMMV (C) + H-or C + B-treated plants at 15, 30, and 45 days after pollination (DAP), respectively. (c) Northern blot and western blot analyses of CGMMV RNA and coat protein (CP) accumulation in watermelon fruits, respectively. The rRNA gel and the Coomassie brilliant blue (CBB)-stained gel are used to show sample loadings. (d-f) Measurements of fruit firmness, mean weight of nine fruits per treatment, and mean endogenous boron content (n = 3) are shown. The results are presented as the mean ± SD. Statistical significance between two different treatments was determined using one-way analysis of variance followed by Duncan's multiple comparison test. Different lowercase letters indicate statistically significant differences between treatments (p < 0.05). Scale bar = 5 cm.

| Foliar application of exogenous boron solution can alleviate CGMMV infection in watermelon fruits
To test the effect of boron application on watermelon resistance to CGMMV infection, we performed four treatments (i.e., phosphatebuffered saline (P) + water (H), P + boron (B), CGMMV (C) + H, and C + B) on watermelon plants (Figure 1a). At 15 days after pollination (DAP), no obvious disease symptoms were observed in the watermelon fruits from all the assayed plants ( Figure 1b)

| CGMMV infection-induced global transcriptomic changes in watermelon fruits
To investigate the effect of exogenous boron application on CGMMV infection, we analysed watermelon fruits from the P + H-, P + B-, C + H-, and C + B-treated plants at 15, 30, and 45 DAP through RNA-Seq. In this study, each treatment had three biological replicates and a total of 36 fruit cDNA libraries were analysed (Table S1). After removal of low-quality reads, a total of 231.44 GB clean reads were obtained with a Q30 percentage >93.08% and a GC content between 43.95% and 44.85% (Table S1). The clean reads were then mapped to the watermelon reference genome using HISAT2 software and resulted in 92.99%-95.94% mapped reads and 90.60%-94.08% unique mapped reads (Table S2). The clean reads data were deposited in the NCBI database (BioProject PRJNA694461). The relative expression levels of individual genes found in each sample were normalized using the FPKM values as the box plots ( Figure S1).  (Figure 2a,b). The results presented in Figure 2c show the numbers of up-or down-regulated genes in each comparison. The results also show that more DEGs were in the C + B vs. P + H, C + B vs. P + B, or C + B vs. C + H comparisons than that in the P + B vs.

| Analysis of DEGs in different groups
P + H comparison at 45 DAP. In addition, more DEGs were in treatment at 45 DAP than that in the same treatment at 15 and 30 DAP.
To further investigate the response in watermelon fruits to CGMMV infection alone and exogenous boron during CGMMV infection, we analysed the DEGs found in the C + H vs. P + H or C + B vs. C + H comparisons at 15, 30, and 45 DAP using the GO terms and the KEGG pathway enrichment. A total of 444 DEGs were found in the C + H vs. P + H comparison and were annotated to GO terms under the categories response to stimulus, immune system process, and cell killing ( Figure 2d and Table S3). The KEGG pathway analysis result indicated that the DEGs found in the C + H vs. P + H comparison were mainly in the categories photosynthesis, carbon metabolism, plant hormone signal transduction, and base excision repair ( Figure 2e and Table S4). A total of 908 DEGs were found in the C + B vs. C + H comparison and were mainly assigned to the GO terms under the categories nucleic acid binding, transcription factor activity, antioxidant activity, and nutrient reservoir activity ( Figure 2f and Table S3). The KEGG pathway analysis result indicated that the DEGs found in the C + B vs. C + H comparison were mainly in the categories phenylalanine metabolism, ascorbate and aldarate metabolism, and biosynthesis of amino acids ( Figure 2g and Table S4). These findings indicate that CGMMV infection can significantly affect the physiological and biochemical responses in watermelon fruits, and foliar application of boron can induce host defence responses as well as antioxidation, transcription factor expression, and nutrition redistribution.

| DEG coexpression network analysis
To investigate the changes of gene regulatory networks in watermelon fruits, we analysed 2043 DEGs using WGCNA and identified 13 distinct gene regulatory network modules containing clusters of highly correlated genes based on pairwise correlation analysis result and gene expression trends. The numbers of DEGs in these modules ranged from 9 to 547 ( Figure 3a and Table S5). The cor- Combination of global WGCNA, topGO, and KEGG analysis results revealed that most of the DEGs identified in this study are involved in plant hormone, signal transport, cell wall, plasma membrane, energy metabolism, and secondary metabolism. In particular, the DEGs in the blue and brown modules are associated with watermelon responses to the C + B treatment.

| Analyses of functions of DEGs involved in polyamine and phytohormone production
Putrescine ( Table S8). The 11 ethylene pathway-associated DEGs were ethylene synthases (ACOs), ethylene receptors (ETRs), and ethyleneresponsive transcription factors (ERFs). Of these ERFs, most were found to be significantly down-regulated in the C + H vs. P + H comparison. Two ACOs were found to be significantly down-regulated, while several ERFs, including RAP2-3 (Cla009283), were found to be significantly up-regulated in the C + B vs. C + H comparison ( Figure 4a,b and Table S8). Moreover, 13 DEGs were found to be related to catabolic processes and abscisic acid (ABA) signal transduction. In the C + B vs. C + H comparison, two ABA 8′-hydroxylase genes were found to be significantly up-regulated. Among the protein phosphatase 2Cs (PP2Cs), the expression of PP2C73 (a negative regulator of ABA signalling) was significantly down-regulated in the C + H vs. P + H comparison (Figure 4a,b and Table S8). Eight DEGs were found to be involved in the synthesis, signalling, and down-  Table S8).
To validate the above findings, we analysed the expression of  Table S8).
To investigate the functions of the polyamine or phytohormone pathway-associated DEGs on WBFD symptom development, we performed VIGS experiments using a CGMMV-based VIGS vector. (h, i) Analyses of polyamines and ABA contents in the fruits from various gene-silenced watermelon plants or controls. The results are expressed as the mean ± SD. The statistical significances were determined using one-way analysis of variance followed by Duncan's multiple comparison test (p < 0.05). Other results were determined using the two-tailed t test. *p < 0.05, **p < 0.01, ***p < 0.001. Scale bar = 5 cm.
Abbreviations of genes and compounds are explained in Table S10.
plants. In addition, an increased (Spd + Spm):Put ratio was found in the fruits from the PP2C73-silenced plants accompanied by a much higher ABA content. In contrast, silencing of SPDS expression resulted in a low ratio of (Spd + Spm):Put and a higher Put content in the fruits (Figure 4h,i).
The above results indicated that (a) polyamine and IAA biogenesis, and ABA catabolism were promoted after the C + B treatment, while ethylene biosynthesis was inhibited; (b) SPDS is required for efficient CGMMV accumulation, while RAP2-3 plays a role in watermelon resistance to CGMMV infection; (c) the ratio of (Spd + Spm):Put is correlated with the relative severity of WBFD in fruits; and (d) the content of Put may have a negative impact on CGMMV infection.

| Roles of DEGs in watermelon antiviral defence signalling
Plants have evolved many defence strategies to resist virus replication and movement (Calil & Fontes, 2017). In this study, we found that most DEGs identified in the C + H vs. P + H or the C + B vs.
C + H comparison were involved in the Ca 2+ signalling pathway (i.e.,  Table S8).
To further validate the above findings, we silenced the expres-

| Roles of DEGs in cell wall and plasma membrane functions, and energy and secondary metabolism
Boron plays important roles in cell wall biosynthesis, and cytoskeleton and membrane functions . In this study, we identified 34 cell wall and plasma membrane-associated DEGs in the C + H vs. P + H and C + B vs. C + H comparisons at 30 and 45 DAP ( Figure 6a and and TDT) are known to be associated with water permeability and pH homeostasis.
The expression of 40 energy and secondary metabolismassociated genes was found to be affected by the C + B treatment ( Figure 6a and Table S8) GSH, and GSSG + GSH (Figure 6i,j). In addition, silencing of PDC1 expression reduced the PDC activity and EtOH content, but increased the contents of PA and LA (Figure 6k).

| The effects of individual silencing homologues of watermelon genes on CGMMV infection in Nicotiana benthamiana
To further explore the roles of watermelon genes in resistance to CGMMV infection, we silenced 12 homologues of the above watermelon genes using the TRV-based VIGS system in N. benthamiana plants (Table S9)

| DISCUSS ION
CGMMV infection-induced WBFD is a devastating disease in watermelon production. In this study, we analysed the function of boron in CGMMV infection in watermelon fruits. Our results show that application of exogenous boron can efficiently inhibit CGMMV infection and minimize WBFD symptoms in watermelon fruits.
Moreover, we have identified through RNA-Seq multiple genes whose expression was regulated by exogenous boron application and/or CGMMV infection. Most of the DEGs identified in this study were found in the 45 DAP samples, suggesting that watermelon resistance to CGMMV infection in the later stage is more complicated.
To validate this speculation, we selected 20 DEGs and analysed their roles in watermelon resistance to CGMMV infection using a CGMMV-based VIGS vector reported previously (Liu et al., 2020).
The results of pV190 VIGS assays for watermelon leaf were consistent with those for fruit except for DnaJ11 ( Figures S2, 4 and 5). In addition, the results of TRV VIGS assays showed that 12 homologues in N. benthamiana played the same antiviral roles as the homologous genes in watermelon ( Figure S3), which further confirmed the reliability of pV190 in functional verification of watermelon fruit genes in resistance to CGMMV infection. We consider that this VIGS procedure can provide researchers with a useful tool for reverse genetics to study gene functions in watermelon and possibly many other cucurbit plants.

| Crosstalk between polyamines and phytohormones can promote or restrict CGMMV infection
Polyamines are essential for plant growth and development and stress responses (Anwar, 2015), including responses to virus infection (Olga et al., 2018).  Table S8). In contrast, the expression of SAMDC, a rate-limiting enzyme that catalyses S-adenosylmethionine (SAM) for Spd and Spm production (Guo et al., 2018), was induced after CGMMV infection and then suppressed after application of exogenous boron (Figure 4a,b and Table S8). These findings indicate that Results are presented as the mean ± SD. The statistical significance between the treatments was determined using one-way analysis of variance followed by Duncan's multiple comparison test (p < 0.05). Other results were analysed using the two-tailed t test. *p < 0.05, **p < 0.01, ***p < 0.001. Scale bar = 5 cm. Abbreviations of genes and compounds are explained in Table S10.

CGMMV infection induces Spd and Spm production in watermelon
fruits. However, after exogenous boron application, watermelon starts to synthesize more Put. This finding supports earlier reports

that impairment of Put biosynthesis increases TRV infection in
Arabidopsis plants (Fernandez-Calvino et al., 2014), and binding of Spd and Spm to DNA and RNA are required during virus replication (Olga et al., 2018). Moreover, a greater ratio of (Spd + Spm):Put is known to promote fruit ripening and softening, and accelerate ABA biosynthesis (Guo et al., 2018), and to alter the expression of PP2C proteins and to regulate ABA production (Anwar, 2015). In this study, we found that more CGMMV accumulated in fruits from the ADC-silenced plants, which had a higher (Spd + Spm):Put ratio and a lower Put content, and in the PP2C73-silenced plants, which had a higher (Spd + Spm):Put ratio and a higher ABA content. In contrast, less CGMMV was detected in the SPDS-silenced plants, which had a lower (Spd + Spm):Put ratio and a higher Put content. Based on the above findings, we propose that knockdown of ADC expression suppresses Put production, and knockdown of PP2C73 expression increases Spd and Spm production, both of which lead to an aggravated virus infection. In addition, a higher (Spd + Spm):Put ratio and a higher ABA content can promote fruit senescence, leading to stronger WBFD symptoms in watermelon fruits. Because silencing of SPDS expression in watermelon plants resulted in a relatively higher Put content in fruits, these fruits accumulated much less CGMMV and showed much weaker WBFD symptoms (Figure 4h,i).
In the crosstalk between polyamines and phytohormones ( Figure 4a), down-regulation of SAMDC expression can inhibit ABA production while promoting IAA production (Guo et al., 2018).
Because SAM serves as a common precursor, there may be an equilibrium between polyamines and ethylene synthesis (Anwar, 2015;Guo et al., 2018). Moreover, boron deficiency and sufficiency can increase and decrease IAA and ABA content, respectively (Eggert & Von Wirén, 2017). In this study, the expression of ethyleneassociated DEGs was found to be down-regulated, while the IAA biosynthesis-associated or ABA catabolism-associated DEGs were found to be up-regulated after C + B treatment (Figure 4a,b and Table S8). We have also found that the boron content was significantly increased in the watermelon fruits from the C + B-treated plants compared to that in the fruits from the C + H-treated plants  Table S8). In contrast, the expression of

| Effect of exogenous boron on other antiviral defence responses
After recognition of pathogens, plants activate various signalling pathways to induce defence responses (Ramirez-Prado et al., 2018). It was reported that calmodulin (CaM), calmodulin-like proteins (CMLs), and calcineurin B-like proteins (CBLs) can convert Ca 2+ signals to induce plant immune responses to bacterial, fungal, and virus invasion (Aldon et al., 2018). CaM is also known to induce SA signalling (Jeon et al., 2017) and SABP2 (an SA receptor) is required for resistance to TMV (Kumar & Klessig, 2003). NIMIN is regulated by SA and can interact with SA sensor protein NPR1 to promote pathogenesis-related protein (PR) gene expression and systemic acquired resistance (SAR) ( Figure 5a) (Hermann et al., 2013). In this study, we have found that the expressions of CaM-1, CMLs, SABP2, and NIMIN-1 were upregulated in the fruits from the C + B treated plants (Figure 5a,b).
Our VIGS assay results showed that silencing of CML48, SABP2,  Table S8). These altered gene expression patterns may also activate host antiviral signalling. Furthermore, the expression of several TFs, especially NACs, MYBs, and WRKYs, were found to be affected by CGMMV infection alone and exogenous boron under CGMMV infection (Figure 5a,b and Table S8). NAC can interact with WRKY, MYB, and TGA to regulate the response to tomato yellow leaf curl virus infection (Huang et al., 2017). In this study, we have found that silencing of NAC2, MYB6, and WRKY12  Table S8).
Numerous RNA viruses cause DNA damage in host cells during their lifecycles (Ryan et al., 2016), H2A is a key component in histone octamers and is responsible for DNA replication and repair (Zhou et al., 2015). Lipid transfer proteins (LTPs) are required for SAR longdistance signalling to resist pepper mosaic mottle virus infection (Sarowar et al., 2009). In our study, the expression of H2A and LTP-DIR1 was down-regulated at 30 DAP and then up-regulated at 45 DAP in the C + B vs. C + H comparison (Figure 5a,b and Table S8). In our VIGS assays, we have found that silencing of H2A or DIR1 expression enhanced CGMMV infection, and silencing of H2A expression also reduced watermelon fruit size (Figures 5d-f and S6), indicating that boron application can affect H2A and DIR1 expression to regulate CGMMV infection. Another gene, SBT, is known to function in pathogen recognition, immune priming, and defence response (Figueiredo et al., 2014). In this study, the expression of SBT was upregulated in the CGMMV-infected fruits, but down-regulated in the fruits from the C + B-treated plants (Figure 5a,b). Silencing of SBT expression through VIGS suppressed CGMMV infection in watermelon fruits but had no effect on fruit weight and quality (Figures 5d-f and S6), suggesting that CGMMV can hijack SBT to ensure its successful infection. Because silencing of SBT expression had no clear effect on fruit weight and quality, we consider SBT to be a candidate in watermelon resistance breeding.

| Cell wall modification-and energy metabolismassociated DEGs enhance host resistance while cytoskeleton-associated DEGs benefit virus infection
The cell wall is a primary barrier to a variety of phytopathogens; viruses, bacteria, and fungi must penetrate it to start the infection process (Smirnova & Kochetov, 2016). Cell wall modification is a crucial element in host defence against biotic stresses (Miedes & Lorences, 2007;Tenhaken, 2015). Xyloglucan endotransglucosylase/hydrolase (XTH) is a hemicellulosic repairing enzyme that can remodel cell walls (Miedes & Lorences, 2007, Tenhaken, 2015. XTH participates in rice resistance to rice stripe virus infection (Zheng et al., 2013). PE/PEI is a cell wall modification enzyme and has two domains (i.e., pectin methylesterase [PME] and pectin methylesterase inhibitor [PMEI]). The PME domain is known to catalyse the demethylation of pectin and to interact with the viral movement protein, while the PMEI domain is known to counteract the actions of plant PMEs to limit virus spread (Lionetti et al., 2014). In this study, the expression of XTH23 and PE/PEI7 was up-regulated in the fruits from the C + B treated plants (Figure 6a and Table S8). Compared with the fruits from the control plants, the fruits from the XTH23-or PE/ PEI7-silenced plants showed much stronger WBFD symptoms and accumulated more CGMMV, but these fruits had much less lignin, cellulose, hemicellulose, soluble pectin, and protopectin (Figure 6ce,g-k), suggesting that boron application can induce XTH23 and PE/ PEI7 expression to strengthen cell walls and to enhance the PMEI function, respectively. In contrast, silencing of XTH23 and PE/PEI7 expression reduced the contents of celluloses, pectins, and lignin, the essential cell wall components (Meents et al., 2018;Verbančič et al., 2018), which destroyed the cell wall barrier, helped viral spread, and aggravated WBFD symptoms.
The cytoskeleton and tubulin are required for virus transport (McLean et al., 1995); boron deprivation can increase the tubulin content in Arabidopsis . In our study, we found that CGMMV infection in watermelon fruits was associated with low boron content (Figure 1f). We also found that at 45 DAP, the expression of TUBB1 was up-regulated in the C + H vs. P + H comparison and down-regulated at 30 DAP in the C + B vs. C + H comparison, indicating that TUBB1 was induced by CGMMV infection and inhibited by exogenous boron (Figure 6a and Table S8). Moreover, the fruits from the TUBB1-silenced plants showed significantly reduced CGMMV accumulation and no WBFD symptoms (Figure 6c-e). These findings agree with previous reports and indicate that TUBB1 has a role in CGMMV infection in watermelon.
In this study, the expression of chloroplast DnaJ11 and ATPS1 was found to be up-regulated in the C + B vs. C + H comparison and down-regulated in the C + H vs. P + H comparison at 45 DAP ( Figure 6a and Table S8). DnaJ proteins are cochaperones that can interact with HSP70s to control protein homeostasis (Bolhassani & Agi, 2019). HSP70s are also known to act as host defence factors or virus help factors to accelerate virus replication and spread in plants (Hýsková et al., 2021). Our VIGS assay result showed that silencing of DnaJ11 expression exacerbated CGMMV infection and WBFD symptom development in watermelon fruits (Figure 6c-e).
We speculate that the up-regulation of DnaJ11 expression through boron application may increase HSP70 activities to enhance watermelon antiviral resistance. ATPS is a key enzyme involved in sulphur assimilation and biosynthesis of GSH and homo-GSH (h-GSH), which are crucial in host stress tolerance (Anjum et al., 2015). In our study, we found that silencing of ATPS1 expression through VIGS increased CGMMV infection and reduced GSSG, GSH, and GSSG + GSH contents in watermelon fruits (Figure 6c-e,i), suggesting that ATPS1 can alleviate WBFD symptoms in watermelon fruits through regulation of sulphur-compound synthesis.
AKR1 has been shown to improve tobacco and rice seed longevity through reducing the contents of cytotoxic compounds (Nisarga et al., 2017). In our study, the expression of AKR1 and  accelerates watermelon fruit decay and acidification (Figure 7).
Taken together, we consider that the results presented in this

| Gene coexpression network analysis
A total of 2043 DEGs in Table S1 were analysed using the WGCNA v. 3.1.1 package in R (Langfelder & Horvath, 2008) to identify the coexpressed gene modules. A gene expression adjacency matrix was constructed to show the network topology with an unsigned type of topological overlap matrix (TOM), a power β of 5, a minModuleSize of 10, and a mergeCutHeight value of 0.25.  (Liu et al., 2020). Briefly, approximately 260 bp fragments representing the partial sequence of one of the 20 selected genes were PCR amplified using the gene-specific primers (Table S9). The resulting PCR products were individually cloned into the pV190 vector using the ClonExpress II One

The constructed plasmids were individually transformed into
Agrobacterium tumefaciens GV3101. Agrobacterium cultures were individually prepared as described (Liu et al., 2020) and diluted to OD 600 = 1.0. The fully expanded cotyledons and the four lowest true leaves of individual 12-leaf-stage watermelon plants (about 1 week before fruit setting) were infiltrated with an Agrobacterium culture containing a specific VIGS vector. The infiltrated plants were grown inside a growth chamber or in a greenhouse for further analyses as described (Liu et al., 2020). Plants infiltrated with an Agrobacterium culture carrying the empty pV190 VIGS vector or a pV190-GFP vector carrying a 266 bp fragment representing the partial sequence of a GFP gene or a pV190-PDS vector carrying a 300 bp fragment representing the partial sequence of watermelon PDS gene were used as controls. Leaf tissues and fruits were collected from these assayed plants at 32 and at 45 dpi, respectively, and were analysed for gene silencing efficiency and CGMMV accumulation.
Twelve genes in N. benthamiana that were homologous with the above watermelon genes were selected to be silenced by TRVbased VIGS vector (Jiao et al., 2020). These approximately 330 bp fragments representing the partial sequence of the 12 genes were amplified individually from an N. benthamiana cDNA through PCR using the PrimeSTAR Max DNA polymerase (TaKaRa) and specific primers (Table S9)

| RT-qPCR
Total RNA was isolated as described above and used for cDNA synthesis using the HiScript III RT SuperMix supplemented with a gDNA wiper (Vazyme). Relative expression levels of the assayed genes and CGMMV RNAs were determined through RT-qPCR using the ChamQ SYBR qPCR Master Mix (Vazyme) on the StepOnePlus Real-Time PCR System (Applied Biosystems). The relative expression level of watermelon Actin gene (Cla007792) was used as an internal control. The RT-qPCR results were calculated using the 2 −ΔΔCt method (Schefe et al., 2006). The primers used in this study are listed in Table S9. Three biological replicates with three technique replicates each were used for each gene.

| Northern blot and western blot analysis
For northern blot assay, total RNA was isolated from mixed watermelon samples. Each mixed sample represented the leaves or fruits from nine watermelon plants, and the RNA samples were separated in 1.5% denaturing agarose gels containing 5% formaldehyde. The separated RNAs were transferred to Hybond-N + nylon membranes (Amersham) through capillary action and the membranes were probed with a digoxigenin-labelled RNA probe, synthesized using a pair of CGMMV-specific primers (CGMMV-F and CGMMV-R) using the DIG Northern Starter Kit as instructed (Roche) ( Table S9). The labelling signal was visualized using the 5200 chemical luminous imaging system (Tanon).
For western blot assay, total protein was extracted from the samples described above using the Plant Protein Extraction Kit (Solarbio), quantified using the bicinchoninic acid (BCA) Protein Assay Kit (TaKaRa), and then separated by 12% SDS-PAGE. The separated proteins were transferred to 0.20 μm polyvinylidene fluoride (PVDF) membranes (Millipore). The resulting membranes were blocked with a bovine serum albumin solution, and then incubated in a 1/4000 diluted anti-CGMMV coat protein monoclonal antibody solution as described (Shang et al., 2011), followed by a 1/5000 diluted goat anti-mouse IgG alkaline phosphatase (AP)conjugated secondary antibody (Sigma-Aldrich). The detection signal was visualized by addition of the chemiluminescent substrate CDP-star (Roche) and then on a 5200 chemical luminous imaging system (Tanon).

| Physiological index measurements
Boron contents in the fruits collected from the four treatments were measured through inductively coupled plasma-mass spectrometry

| Statistical analysis
The RT-qPCR and biological index measurement results were further analysed using one-way analysis of variance followed by Duncan's multiple comparison test. The VIGS assay results were analysed using Student's two-tailed t test. The statistical significances were determined using SPSS Statistics 25 (IBM Inc.).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found in the online version of the article at the publisher's website.