Transcriptomic and metabolomics responses to elevated cell wall invertase activity during tomato fruit set

Ovaries are transcriptionally much more sensitive to metabolic intervention than fruitlets and elevation of CWIN activity increased the expression of R and PR genes in ovaries for defence against pathogens.


Introduction
Fruit set is a critical process determining yield potential of many crops. Most studies in fruit biology, however, have focused on mid-to-late stages of development, especially in relation to fruit expansion (Carrari and Fernie, 2006), sugar accumulation (Beauvoit et al., 2014) and ripening (Alba et al., 2005). Only a limited number of studies have explored the molecular regulation of fruit set, characterized by the Ruan 2012). A classic example is from the miniature 1 mutant in maize, which exhibits a miniature seed phenotype due to a mutation in the Mn1 gene encoding a CWIN, INCW2, leading to reduced CWIN activity and cell division (Cheng et al., 1996). In rice, GIF1, encoding a CWIN, is a key regulator of grain filling. Overexpressing GIF1, driven by its native promoter, resulted in larger and heavier grains in the transgenic line (Wang et al., 2008). Silencing of a CWIN gene, Lin5, in tomato led to fruit abortion (Zanor et al., 2009), while elevated CWIN activity promoted fruit and seed development (Jin et al., 2009;Liu et al., 2016). The hexoses produced from CWIN-mediated hydrolysis of Suc unloaded to the ovary/ fruitlet apoplasm could function both as signalling molecules to regulate development and as an energy source and building blocks for metabolism and growth. Ruan et al. (2012) proposed a model of sugar-mediated seed and fruit seed set, in which a Glc signal, produced by CWIN, is hypothesized to promote cell division and to suppress the programmed cell death pathway for successful fruit set. However, it remains unknown which molecular pathways underlie CWINmediated fruit set.
We aimed to identify which genes and molecular pathways are responsive to elevated CWIN activity during tomato fruit set, focusing on the ovary-to-fruit transition period of 2 d before anthesis (dba) ovaries to 2 d after anthesis (daa) fruitlets. To this end, LIN5 and SlINVINH1 are the only CWIN and CWIN inhibitor genes, respectively, expressed in tomato ovaries and fruits (Jin et al., 2009;Palmer et al., 2015). Silencing SlINVINH1 elevated CWIN activity by ~ 35% in 2 dba ovaries and 2 daa fruitlets (Liu et al., 2016). Unlike some ectopic expression studies, where misexpression of an introduced foreign gene often causes pleiotropic and detrimental effects on growth and development, no foreign CWIN genes are introduced in this case and the elevation of CWIN activity was derived entirely from the endogenous CWIN enzyme (Jin et al., 2009). This provides a valuable opportunity to assess the role of the native CWIN in tomato fruit set without potential complications from misexpression of foreign CWIN genes.
We identified possible molecular pathways responsive to elevated CWIN activity by using transcriptomic and metabolomic approaches. The analyses revealed a number of novel findings including the following: (i) ovaries were much more responsive to elevated CWIN activity as compared with fruitlets, and (ii) elevated CWIN activity enhanced the expression of genes involved in pathogen resistance, ethylene synthesis, cell cycle and cell wall synthesis/remodelling, but reduced the expression of genes associated with photosynthesis and protein degradation as well as those encoding some receptor like kinases (RLKs). Overall, our data provide important insights into the molecular pathways by which CWIN regulates fruit set.

Plant material
Transgenic tomato plants (Solanum lycopersicum XF-2) were used in which CWIN activity was elevated by RNAi silencing of its inhibitor gene, SlINVINH1 (Jin et al., 2009). The transgenic and WT plants were grown in a glasshouse under natural light, with day and night temperatures of 25 and 18 °C for 14 and 10 h, respectively. Plants were raised in 25-cm diameter (8 litres) pots filled with potting mix (1 part coarse sand, 1 part perlite and 1 part coir-peat), with one plant per pot. Standard Osmocote™/Osmocote high (K) potassium™ 1:1 slow release fertilizer (Scotts) was applied at a rate of 20 g per pot, supplemented with a weekly liquid fertilizer regime of Jurox Wuxal Liquid Foliar Nutrient Fertilizer™ at a diluted concentration of 4 ml l −1 . Pots were maintained at field capacity through being watered twice a day, each of 3 min duration, by an automated drip irrigation system.
Tomato fruit age was determined by tagging flowers on the day of anthesis. To ensure synchronous pollination, each flower of interest was manually pollinated using a battery-powered vibrator at 10:00-11:00 h on the day of anthesis. This technique has proven to be highly effective in releasing sufficient pollen grains for synchronous pollination and fruit set (e.g. Palmer et al. 2015;Liu et al., 2016). Ovaries and fruitlets at 2 dba and 2 daa, respectively, were harvested and frozen in liquid nitrogen immediately and then stored at −80 °C until use.

RNA isolation and Illumina sequencing
Total RNA was extracted from 2 dba ovaries and 2 daa fruitlets of WT and SlINVINH1-RNAi plants using Qiagen RNA Plant Mini Kits. Genomic DNA was removed by using DNase (Promega). A 5 μg subsample of total RNA from each of 12 RNA extracts (2 stages×2 genotypes×3 biological replicates) were sent to the Australia Genome Research Facility Ltd (AGRF) in Melbourne, Australia. The Illumina Hiseq-2000 platform was used to produce 100 bp single-end runs.
Total RNA quality was determined using an Agilent 2100 Bioanalyzer and quality of the sequenced raw reads were assessed by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/ fastqc/).
Principal component analyses CLC Genomics Workbench (9.0) was used for principal component analysis (PCA) of RNA-seq and metabolomic data sets.
qRT-PCR RNA extraction and qRT-PCR were carried out as described previously (Palmer et al., 2015). SlCAC and SlTIP41 were used as reference genes as the expression levels of these genes were most stable from the ovary-to-fruitlet stages in both RNAi and WT plants. Primers used for qRT-PCR were list in Supplementary Table S1 at JXB online.

Identification of differentially expressed genes during fruit set in SlINVINH1-RNAi and WT plants
As the first step, we confirmed that the CWIN activity was elevated in ovaries and fruitlets of the SlINVINH1-RNAi plant. Enzyme assays revealed that CWIN activity was indeed significantly increased in the 2 dba ovaries and 2 daa fruitlets compared with that in WT plants (see Supplementary  Fig. S1), consistent with our recent report (Liu et al., 2016). Importantly, the activities of vacuolar invertase (VIN) and cytosolic invertase (CIN) were not affected in the transgenic plants ( Supplementary Fig. S1), demonstrating the specific inhibitory effect of SlINVINH1 against CWIN as shown previously (Jin et al., 2009). Following this validation, we conducted RNAseq analyses to discover which genes or gene networks responded to the elevated CWIN activity in the RNAi plants.
Among the total 34 727 genes in the tomato genome, our RNAseq analysis indicated that about 58% of them were expressed in the 2 dba ovaries and 2 daa fruitlets with little difference between the two developmental stages or between RNAi and WT plants. We then identified differentially expressed genes (DEGs) at fruit set of WT and SlINVINH1-RNA interference (RNAi) plants. In total, 5183 and 3262 DEGs were detected in WT and SlINVINH1-RNAi plants, respectively, during the ovary-to-fruit transition (Fig. 1A, Supplementary Tables S2 and S3). These represent 15.0% and 9.4% of the genes in the tomato genome, respectively. It is intriguing that DEG numbers were much less in SlINVINH1-RNAi compared with WT plants during fruit set (Fig. 1A). The number of DEGs (5183) detected during WT fruit set was four times higher than those reported by Wang et al., (2009) who identified 1298 DEGs during tomato fruit set from 2 dba to 4 daa. These authors used a microarray approach that does not provide genome-wide coverage. The RNA sequencing applied in our study overcame this obstacle, leading to the detection of many more expressed genes.
Functional categorization, using MAPMAN, of the DEGs (5183) in the WT samples revealed that major groups exhibiting changes during fruit set belong to categories for protein (BIN 29), RNA (BIN 27), and transport (BIN 34). The protein (BIN 29) category consisted of 14.7% of the total DEGs during WT fruit set ( Table 1). The protein category (BIN 29) contains protein synthesis (BIN 29.2), protein post-translational modification (BIN 29.4), and protein degradation (BIN 29.5) (Thimm et al., 2004). RNA (BIN 27) was the second major group, containing 12.9% of the total DEGs in WT fruit set (Table 1) and can be subdivided into RNA processing (BIN 27.1), RNA translation (BIN 27.2), regulation of transcription (transcription factor, BIN 27.3), and RNA binding (BIN 27.4) (Thimm et al., 2004). A similar grouping of functional categories to WT fruit set was found during fruit set of the SlINVINH1-RNAi line, but with a lower number of DEGs in each of these functional groups. The top two functional groups were protein (BIN 29) and RNA (BIN 27). The ranking of other functional groups was slightly altered in the SlINVINH1-RNAi compared with WT fruit set ( Table 1). The large number of genes involved in protein (BIN 29) and RNA (BIN 27) in both genotypes indicates that fruit set is a process undergoing rapid cell development with intensive protein and RNA synthesis, regulation and turnover.
Photosynthesis-related genes were predominately down-regulated during fruit set among the 497 DEGs unique to elevation of CWIN activity To unravel the candidate genes that are regulated by the elevation of CWIN activity during fruit set process, we prepared a Venn diagram to show how many DEGs are unique to fruit set of SlINVINH1-RNAi plants. This analysis revealed that the majority of DEGs (2765) were common to both genotypes during fruit set, with only 497 DEGs uniquely responsive to elevated CWIN activity ( Fig. 1B and Supplementary  Table S4).
Of particular note, an interesting expression pattern of photosynthesis-related genes was observed among the 497 DEGs. With the exception for the gene encoding rubisco activase, which was up-regulated, expression levels of the other 13 of 14 DEGs encoding photosynthesis proteins were decreased in 2 daa fruitlets compared with 2 dba ovaries of SlINVINH1-RNAi plants (Table 2). These photosynthesisrelated genes encode proteins functioning in different photosynthetic processes. Among them, nine transcripts encode proteins for the light reaction, one for photorespiration and four for the Calvin-Benson cycle. For those responsible for the light reactions, three transcripts encode chlorophyll a/b binding proteins (Table 2).
Consistent with the repression of a large number of photosynthesis genes, developing fruits of the transgenic plants were paler green at 10 and 15 daa compared with WT fruits (Fig. 2) although such a phenotype was not apparent for 2 dba ovaries or 2 daa fruitlets.  Differentially expressed genes were ascribed to 36 bins in MAPMAN (Thimm et al., 2004). Only the major functional groups are listed. The order was made based on the number of differentially expressed genes in a given functional category (BIN), excluding the categories of 'Others' and 'Unknown'. Ovaries were transcriptionally much more sensitive to elevated CWIN activity than fruitlets We then compared the impact of elevated CWIN activity on the transcriptomes of 2 dba ovaries and 2 daa fruitlets between WT and SlINVINH1-RNAi plants. The response of ovaries and fruitlets to elevated CWIN activity was remarkably different. There were only seven DEGs found in 2 daa fruitlets between the two genotypes. However, surprisingly there were 319 DEGs observed in 2 dba ovaries between the two genotypes (Fig. 3). These data show that 2 dba ovaries were much more responsive to elevated CWIN activity than 2 daa fruitlets at the gene transcript level.
Among the seven DEGs in 2 daa fruitlets ( Fig. 3), four transcripts belong to categories encoding proteins functioning in the cell wall, signalling, biotic stress, and lipid metabolism, with a further three transcripts encoding proteins of unknown function (Table 3). The four DEGs of known function encode a pectin esterase, a leucine-rich repeat (LRR) receptor-like serine/threonine-protein kinase, a resistance (R) protein containing a nucleotide binding site and leucinerich repeat (NBS-IRR) domains, and cyclopropane-fattyacyl-phospholipid synthase. Expression levels of six DEGs decreased in 2 daa fruitlets of SlINVINH1-RNAi compared with WT fruitlets, except for transcript levels of a gene encoding an R protein which increased in 2 daa fruitlets of SlINVINH1-RNAi compared with WT fruitlets (Table 3).
For the 319 DEGs detected in the genotypic comparison of 2 dba ovaries (see Supplementary Table S5), a functional categorization analysis revealed that DEGs encoded proteins primarily functioning in protein (BIN 29), hormone (BIN 17), RNA (BIN 27), stress (BIN 20), signalling (BIN 30), and cell wall (BIN 10) (Table 4). Since ovaries, not fruitlets, were the primary organ responsive to elevated CWIN activity in the SlINVINH1-RNAi plants, we focused our subsequent analyses on 2 dba ovaries in order to gain further insights into the DEGs regulated by increases in CWIN activity.

Protein degradation-related genes were largely repressed in ovaries of SlINVINH1-RNAi plants in response to elevated CWIN activity
We detected 24 DEGs involved in protein degradation in ovaries of SlINVINH1-RNAi compared with WT plants (Table 5). These DEGs encode various proteases including subtilases, members of the subtilisin-like serine protease family (Siezen and Leunissen, 1997). All six DEGs encoding subtilisin-like proteases exhibited decreased transcript levels in ovaries of SlINVINH1-RNAi compared with WT plants. However, DEGs encoding cysteine protease, aspartate protease, and serine protease proteins, as well as those involved in the ubiquitin-proteasome pathways, exhibited either up-or down-regulation (Table 5). Overall ~70% (16 out of 24) of DEGs encoding proteins functioning in protein degradation exhibited a decreased expression in ovaries of SlINVINH1-RNAi compared with those of WT plants, indicating that protein degradation may be repressed in ovaries of SlINVINH1-RNAi compared with WT ovaries ( Table 5).

SlINVINH1-RNAi plants in response to elevated CWIN activity
Considering the important roles hormones play in fruit set (de Jong et al., 2009), and sugar-hormone interactions (Leon and Sheen, 2003;Kumar et al., 2014), we closely examined hormone-related transcripts.
During fruit set in WT plants, genes involved in auxin, GA, and ethylene metabolism and signalling are among the top three 'hormones' gene classes, representing 27%, 17%, and 17% of the total hormone-related genes respectively (Fig. 4).  A similar expression profile was found in the SlINVINH1-RNAi plants, with genes involved in ethylene synthesis and signalling ranked the third most important hormone class in terms of DEG numbers (Fig. 4). These findings indicate that ethylene may also be an important player during fruit set, in parallel to the two well-known players, auxin and GA.
The last step of the ethylene biosynthetic pathway, 1-aminocyclopropane-1-carboxylate oxidase (ACC oxidase) catalyses the conversion of aminocyclopropane-1-carboxylate (ACC) to ethylene (Yang and Hoffman, 1984). During WT fruit set, expression levels of four DEGs encoding ACC oxidase proteins decreased while the other 15 DEGs increased in 2 daa fruitlets compared with 2 dba ovaries (Table 6). Similarly, in the SlINVINH1-RNAi plants, four ACC oxidase transcript levels decreased whereas nine transcript levels increased in 2 daa fruitlets compared with 2 dba ovaries ( Table 6). The results suggest that ethylene synthesis capacity may be higher in 2 daa fruitlets than 2 dba ovaries in both WT and SlINVINH1-RNAi plants. Compared with WT ovaries, four ACC oxidase transcripts exhibited up-regulation with one down-regulated in ovaries of SlINVINH1-RNAi plants ( Table 6), suggesting that ethylene synthesis may be enhanced in the CWIN-elevated ovaries.
In addition to ethylene synthesis genes, expression of a gene encoding an ethylene response factor (ERF) was upregulated 2.83-fold in SlINVINH1-RNAi ovaries compared with WT ovaries (Table 5). Transcripts encoding ERF3 and ERF2a were also up-regulated by 2.26-and 0.78-fold, respectively, in 2 dba ovaries of SlINVINH1-RNAi compared with those of WT plants. These two ERFs are annotated as pathogenesis-related (PR) transcriptional factors (Table 5), which belong to the APETALA2/ethylene-responsive element binding protein family, and have been demonstrated to be able to bind to the promoters of several PR genes (Singh et al., 2002). The increased expression of most transcripts encoding ACC oxidase proteins and up-regulation of three ERFs imply that ethylene production and sensitivity to ethylene are most likely higher in 2 dba ovaries of SlINVINH1-RNAi compared with those of WT plants.

Defence related R genes were up-regulated in ovaries of SlINVINH1-RNAi plants in response to elevated CWIN activity
CWIN is involved in stress responses, including biotic stress (Ruan, 2014) and abiotic stress (McLaughlin and Boyer, 2004;Liu et al., 2013). During biotic stress, R genes encoding proteins containing NBS-IRR domains are often expressed to interact with pathogen effector proteins. Plants have evolved hundreds of R genes to defend themselves against pathogens by activating defence response, including PR gene expression (Lee and Yeom, 2015). In this section, we focus on the expression of stress-related genes, R and PR gene, during fruit set and their response to elevated CWIN activity.
Five out of six DEGs encoding R proteins were up-regulated in 2 dba ovaries of SlINVINH1-RNAi compared with those of Table 4. Functional classification of differentially expressed genes in 2 dba ovaries of SlINVINH1-RNAi compared with those of wild type (WT) plants Differentially expressed genes were ascribed to 36 bins in MAPMAN (Thimm et al., 2004). Some genes might be allocated to multiple different bins.   WT plants (Table 5). Among them, three R genes contain Toll/ interleukin-1 receptor (TIR) motif with one R gene containing the coiled-coil (CC) motif in the N-terminal region; the other two R genes may lack the N-terminal motif (Table 5). Apart from the R genes, expression of two genes encoding endochitinase were altered in 2 dba ovaries of SlINVINH1-RNAi compared with WT plants (Table 5). Endochitinase belongs to the PR protein family (Ebrahim et al., 2011). Together, elevated CWIN activity not only enhanced expression of R genes but also altered the expression of a PR gene in 2 dba ovaries of SlINVINH1-RNAi plants compared with those of WT plants.

Cell wall-related genes in ovaries of SlINVINH1-RNAi plants in response to elevated CWIN activity
Cell wall metabolism has been well studied in the later stages of fruit development, namely, ripening and senescence (Brummell et al., 1999), but not in the early fruit set stage. Cell walls not only provide mechanical support to plants but also function as protective barriers against pathogen attack and a pathway for apoplasmic transport of resources including water, ions, and sugars (Palmer et al., 2015). There were 13 DEGs encoding proteins involved in cell wall metabolism in 2 dba ovaries of SlINVINH1-RNAi compared with WT plants (Table 5), including cellulose synthase, endo-1,4-β-glucanase, β-xylosidase, xyloglucan endotransglucosylase/hydrolase 2, pectin esterase, pectin acetylesteresteras, and expansin ( Table 5). The expression changes in these cell wallrelated genes indicated that cell wall metabolism was affected in response to elevated CWIN activity in 2 dba ovaries of SlINVINH1-RNAi compared with those of WT plants.

RLK-related genes were largely down-regulated in ovaries of SlINVINH1-RNAi plants in response to elevated CWIN activity
Signalling-related genes also were targeted in our analyses, since Glc or Fru produced from CWIN activity in the extracellular space may be sensed by as yet unidentified receptors on the plasma membranes to trigger downstream signal transduction (Ruan, 2014).
There were 13 DEGs encoding RLK proteins, with 11 exhibiting decreased expression levels in 2 dba ovaries of SlINVINH1-RNAi compared with those in WT plants. These included two DEGs encoding cytoplasmic RLKs, which also exhibited reduced expression in the transgenic ovaries (Table 5). Unlike other RLKs, which are predicted to localize to the plasma membrane, these two receptor-like cytoplasmic kinases lack an extracellular domain, leading to their cytoplasmic localization (Afzal et al., 2008). Interestingly, expression of a small GTPase gene (Rho of plants; ROP) also was down-regulated in the SlINVINH1-RNAi ovaries compared with WT (Table 5).

Cell cycle genes were generally up-regulated in ovaries of SlINVINH1-RNAi plants in response to elevated CWIN activity
Fruit set is characterized by intensive cell division (Ruan, 2012), a process stimulated by Glc signalling that could originate from CWIN activity (Wang and Ruan, 2013). Therefore, we examined expression of cell division-related genes to see how they responded to elevated CWIN activity. We identified two DEGs encoding cyclin B2 in 2 dba ovaries of SlINVINH1-RNAi plants (Table 5), which were up-regulated in 2 dba ovaries of SlINVINH1-RNAi compared with that in WT plants (Table 5).

Validation of RNA-seq data by qRT-PCR
To validate the RNA-seq data, we selected six DEGs from Table 5 for qRT-PCR measurements of their respective transcripts from 2 dba ovaries of the RNAi and WT plants. The analyses revealed that all the six genes exhibited the same trends in their RNA fold changes between the two genotypes measured by both methods. We noticed that the fold changes measured by qRT-PCR were generally lower than that from  Molecular pathways of ovary-to-fruit transition | 4271 RNA-seq analyses, probably owing to the differences in the sensitivity and specificity between the two approaches (Griffith et al., 2010). Importantly, the regression coefficient between the two datasets is 0.7144 (see Supplementary  Fig. S2), which is within the range of reported correlations between RNA-seq and qRT-PCR results (e.g. Xu et al., 2011). Together, the findings indicate the reliability of our RNA-seq data.

Primary metabolite profiling during fruit set
To understand how the CWIN-mediated changes in transcript profiling as described above may impact on metabolism during fruit set, we undertook a metabolomic analysis for the same set of samples used for RNAseq. In total, 55 metabolites were measured and these were subdivided into four groups: (i) amino acids; (ii) sugars and sugar alcohols; (iii) organic acids; and (iv) miscellaneous (Table 7). Specifically, 24 amino acids, 16 sugars and sugar alcohols, 12 organic acids and three miscellaneous metabolites were detected. Among the 24 amino acids, 22 exhibited slightly reduced levels in ovaries of the RNAi plants compared with those in the WT, although only two, isoleucine and tryptophan, showed a statistically significant decrease (Table 7). During the ovary-to-fruitlet transition, six amino acids were significantly increased in WT plants with a similar trend observed for the transgenic plants. Noticeably, levels of methionine, the precursor for ethylene biosynthesis, were doubled during the ovary-to-fruit transition in both genotypes (Table 7).
For the sugar and sugar alcohol group, most members did not show significant changes in the RNAi ovaries compared with those of the WT, except for fructose, fucose, and isomaltulose, which exhibited significant reductions (Table 7). Interestingly, both G-6-P and F-6-P exhibited significantly increased levels across the ovary-to-fruit transition in both WT and SlINVINH1-RNAi plants but without being effected by elevated CWIN activity in the transgenic ovaries (Table 7).
Other metabolites exhibiting significantly increased levels in both genotypes during the ovary-to-fruit transition included maltotriose, succinic acid, nicotinamide, nicotinic acid, and glucuronic acid. Among these metabolites it is interesting to note that the levels of glucuronic acid, succinic acid, and nicotinamide were significantly increased in the CWIN-elevated transgenic ovaries and fruitlets as compared with their respective WT counterparts (Table 7).

Principal component analyses of transcriptome and metabolome data
Principal component analyses (PCA) of transcriptome and metabolome data allowed clear visualization of differences between developmental stages of fruit set and genotypes studied. Each point represented the transcriptome or metabolome of the different samples, in a two-dimensional plot. At the ovary stage, where we observed significant transcriptomic shifts in response to elevation of CWIN (Fig. 3), both transcriptome and metabolome were clearly separated between the CWIN-elevated and WT plants (Table 7). At the fruitlet stage (2 daa), however, transcriptome differences could not be distinguished between the two genotypes. By contrast, they separated well into two distinct clusters using the metabolomics data (Fig. 5). We attribute this finding to differences in CWIN activity being more discriminative at the metabolomic compared with the transcriptomic level and to metabolic changes preceding changes at the transcriptional level in the fruitlets. A similar phenomenon was also observed in the study of Wang et al. (2009) who compared auxin response factor IAA9 in transgenic and WT tomato plants and found that the difference in the metabolomic data is more discriminative than that in the transcriptomic data in their comparisons.

Discussion
Ovaries are transcriptionally far more sensitive to elevated CWIN activity than fruitlets: ready to set Upon fertilization, tomato ovaries develop into fruitlets. However, the molecular pathways responsible for ovary-tofruit transition and how they respond to metabolic or biochemical changes remain largely unknown. Our analyses revealed that only seven transcripts exhibited changed levels in response to elevated CWIN activity at the fruitlet stage. By contrast, 319 transcripts were differentially expressed in ovaries of SlINVINH1-RNAi (CWIN activity elevated) compared with WT ovaries (Fig. 3), a 45-fold increase in the number of DEGs compared with the fruitlet stage. These findings indicate that ovaries undergo a much more complex regulation in response to elevated CWIN activity at the mRNA level than fruitlets, in order to ensure successful pollination and fertilization.
To the best of our knowledge, this is the first report illustrating that ovaries were much more responsive to elevated CWIN activity compared with fruitlets. Indeed, there are no studies reporting that the ovary is more sensitive to changes in gene expression compared with fruitlets. With this information, we focused on the ovary stage for detailed comparative analyses. In broad terms, this finding is in agreement with our 'Ready-Set-Growth' model of fruit set, in which significant molecular and biochemical changes are proposed to take place in the ovary preceding transition to the fruitlet stage upon fertilization Palmer et al., 2015). Consistently, a recent study showed that a group of CWIN and VIN genes in maternal reproductive tissues of tobacco flowers were induced by pollination (Goetz et al., 2017).

Elevation of CWIN activity increased the expression of genes for defence against pathogens
During fruit set in WT plants, expression levels of most genes encoding R proteins involved in pathogen resistance were higher in 2 dba ovaries than that in 2 daa fruitlets (see Supplementary  Table S2). Vriezen et al., (2008) found that expression of genes encoding defence-related proteins decreased from 3 dba ovaries to 3 daa fruitlets in tomato. Our observation, together with that reported by (Vriezen et al., 2008), indicates that ovaries are likely to be under a default active state against potential pathogen attack, probably as a mechanism to protect the critical transition from ovules to seeds and ovaries to fruitlets. The higher expression levels of R genes in 2 dba ovaries compared with 2 daa fruitlets in WT plants (Table 5) also suggests that ovaries are more prone to pathogen attack than fruitlets. Indeed, ovaries are often under attack by many pathogenic Table 7. Primary metabolites levels in ovary (2 dba) and fruitlets (2 daa) of the wild type (WT) and SlINVINH1-RNAi line Data are normalized to mean values of the wild type ovaries (2 dba). Values are means±SE of three replicates. Bold values indicate significant differences by Student's t-test (P<0.01) of the WT fruitlets (2 daa) and SlINVINH1-RNAi line compared with the WT ovary (2 dba). Asterisk indicates significant differences by Student's t-test (P<0.01) of the SlINVINH1-RNAi line fruitlets (2 daa) in comparison with the same line at ovary stage (2 dba).

Metabolite
WT ovaries (2 dba) WT fruitlets (2 daa) SlINVINH1-RNAi ovaries (2 dba) fungi, some of which can directly penetrate ovary walls (Ngugi and Scherm, 2006). In addition, tomato fruit can be infected by pathogens such as Xanthomonas campestris pv. vesicatoria, which causes a serious disease (Oldroyd and Staskawicz, 1998). The R gene-mediated pathway could contribute a broad spectrum of resistance, including resistance against X. c. vesicatoria (Oldroyd and Staskawicz, 1998). Six R genes and two PR genes were responsive to elevated CWIN activity in 2 dba ovaries of SlINVINH1-RNAi in comparison with those in WT plants (Table 5). Among them, five R genes exhibited up-regulated expression in transgenic ovaries with one PR gene up-and one down-regulated. These findings indicated that elevation of CWIN activity enhanced overall R gene expression in ovaries, and hence the potential of the transgenic ovaries to defend against pathogen attack. PR gene expression is part of the R gene-mediated defence reaction in plants (Lee and Yeom, 2015). Enhanced PR gene expression by expression of invertases has been reported previously (Herbers et al., 1996;Sonnewald et al., 2012;Sun et al., 2014).

SlINVINH1-RNAi fruitlets (2 daa)
How CWIN activity induces expression of PR genes remains unclear. In this context, transgenic tobacco plants expressing a yeast-derived CWIN or VIN gene developed necrotic lesions in source leaves, typical of a hypersensitive responses (HR) symptom, with three transcripts (PR-Q, PAR1, PR-1b) encoding PR proteins accumulated in the transgenic tobacco leaves (Herbers et al., 1996(Herbers et al., , 2000. Similar observations have been made recently in source leaves of pepper (Sonnewald et al., 2012) and rice (Sun et al., 2014). These authors proposed that invertase-mediated sugars induced PR gene expression via the secretory pathway through which CWIN and VIN proteins are targeted to the cell wall matrix and vacuoles, respectively. Indeed, CWINs are also considered as PR proteins in their own right (Roitsch et al., 2003). These studies, however, did not report induction of R genes by increased CWIN activity (von Schaewen et al., 1990;Herbers et al., 1996). Our data indicate that the CWIN induces or enhances R gene expression, leading to activation of PR genes.
There are a few partners in the R-gene-mediated pathway for resistance to pathogen infection. One of them is cytoplasmic serine/threonine kinase. In addition to the increased R gene expression in 2 dba ovaries of SlINVINH1-RNAi

Metabolite
WT ovaries (2 dba) WT fruitlets (2 daa) SlINVINH1-RNAi ovaries (2 dba)  Table 7. Continued compared with WT ovaries (Table 5), we also found two DEGs encoding cytoplasmic serine/threonine kinases in the SlINVINH1-RNAi ovaries (Table 5). A typical example of the interaction of cytoplasmic serine/threonine kinase and R protein is the Prf-Pto complex in tomato pathogen recognition (Ntoukakis et al., 2014). Prf is an R gene, containing a leucine-rich repeat and nucleotide-binding site motif (Salmeron et al., 1996), with the encoded R protein localized to the cytoplasm (Gururani et al., 2012). Pto encodes serine/threonine protein kinase without an extracellular domain, also located in the cytoplasm (Martin et al., 1993;Salmeron et al., 1996). Pto can bind directly to the N-terminus of Prf in vivo (Mucyn et al., 2006). However, the exact mechanism of how the Prf-Pto complex works in recognizing effectors secreted by pathogens is unknown (Oh and Martin, 2011). Identification of two DEGs encoding cytoplasmic serine/threonine kinases in the CWIN-elevated tomato ovaries provided an opportunity to examine how alterations in Suc catabolism and signalling may modulate their interaction with R proteins to respond to pathogen infection in tomato. ERFs, such as Pto-interacting 4 (pti4), are regarded as another partner of the Prf-Pto complex (Wu et al., 2015). An ERF physically interacts with Pto and activates PR gene expression (Zhou et al., 1997;Chakravarthy et al., 2003). The two ERFs in our study were annotated as PR transcriptional factors, and thus may represent other partners in the R-gene mediated pathway.
ROP, a small GTP-binding protein, is also considered to be involved in interaction with the Prf-Pto complex and has a role in ROS production, leading to the HR reaction (Kawano et al., 2010). Interestingly, a ROP gene was downregulated in SlINVINH1-RNAi compared with WT ovaries (Table 5). Apart from its role in HR response, ROP functions as a signalling switch that controls a wide variety of cellular functions including cell division, cell differentiation, cell morphogenesis (Yang, 2002;Miyawaki and Yang, 2014), and plant defence responses (Agrawal et al., 2003;van Zanten et al., 2009;Shpak, 2013). The down-regulation of ROP gene expression in tomato ovaries in response to elevated CWIN activity indicates that CWIN may regulate fruit set by affecting a broad array of cellular development pathways.
Together, the above findings provide opportunities to further study the molecular mechanisms by which CWINmediated sugar signalling regulates pathogen resistance in ovaries and fruits through the R-gene mediated pathway, which may involve the expression and function of cytoplasmic serine/threonine kinase, PR ERFs and ROP.
Attention was also given to the possible trade-off for nutrients and energy between plant defence and development (Messina et al., 2002;Huot et al., 2014). Here, levels of most amino acids displayed a slight reduction in 2 dba ovaries of the CWIN-elevated plants compared with those of WT with the difference being statistically significant for isoleucine and tryptophan (Table 7). The reduced size of the free amino acid pool may be derived from decreased expression of protein degradation genes in the transgenic ovaries (Table 5). Alternatively, it may be indicative of enhanced protein synthesis including that for the R proteins since amino acids are building blocks for protein synthesis. The protein content was indeed increased in 10-day-old seeds of SlINVINH1-RNAi plants compared with WT plants (Jin et al., 2009). Protein synthesis consumes substantial metabolic energy derived from Glc, which might have impacted fruit development. However, fruit set was not affected under optimal condition and was enhanced under heat stress in SlINVINH1-RNAi plants compared with WT plants (Liu et al., 2016), indicating that the energy status was not adversely affected in transgenic ovaries or fruitlets. To this end, expression of most protein degradation-related genes were downregulated in 2 dba ovaries of SlINVINH1-RNA compared with WT ovaries (Table 5). This indicates a prolonged protein turnover rate, which would save metabolic energy for synthesis of new proteins including those for defence-related functions in the CWIN-elevated ovaries to increase their defence capacity for fruit set. Synergistically, enhanced CWIN activity also improved fruit set under long term moderate heat stress (Liu et al., 2016). Thus, elevation of CWIN activity appears to enhance the defence capability of ovaries and fruitlets against both biotic and abiotic stress.

Elevation of CWIN activity repressed photosynthesisrelated genes during ovary-to-fruitlet transition
During ovary-to-fruit transition, we found that the CWINelevated plants exhibited decreased transcript levels of a large number of photosynthetic genes controlling both light and dark reactions as compared with the same ovary-to-fruit transition stage in WT plants (Table 2). Consistently, the transgenic fruits displayed a lighter colour late in development (Fig. 2). Repression of photosynthesis genes by expressing a yeast-derived invertase in the apoplast has been reported in mature leaves of tobacco and potato (Heineke et al., 1992). The effect is thought to be caused by a feedback inhibition of Suc export from source leaves (von Schaewen et al., 1990;Dickinson et al., 1991) due to Suc being hydrolysed by CWIN into Glc and Fru, which cannot be loaded into phloem for translocation to sinks organs (von Schaewen et al., 1990). Moreover, high levels of Glc, could inhibit photosynthesis directly though sugar signalling (Jang and Sheen, 1994;Sheen et al., 1999). With higher CWIN activity in the transgenic ovaries and fruitlets, extra Glc is likely to be released from Suc hydrolysis in their apoplasms, which may repress photosynthesis through a feedback system. Physiologically, enhanced assimilate import observed in the RNAi fruit (Liu et al., 2016) may make it unnecessary for fruit to produce extra carbon via photosynthesis. Indeed, a study in which chlorophyll production was down-regulated in tomato revealed that fruit photosynthesis was non-essential for normal fruit function but required for seed set (Lytovchenko et al., 2011).

Elevation of CWIN activity enhanced expression of ethylene synthesis and cell cycle genes
Genes involved in ethylene synthesis and signalling were ranked third within the hormone-related gene category in terms of number of DEGs after auxin-and GA-related genes expressed during ovary-to-fruit transition (Fig. 4). Previous studies also found that expression of ethylene-related genes was among the most altered across of all hormone-related genes during tomato fruit set (Vriezen et al., 2008;Wang et al., 2009). Auxin and GA are well-recognized hormones during fruit set (Goetz et al., 2007;de Jong et al., 2009). Ethylene may play an antagonistic role to auxin and GA (Vriezen et al. (2008). It is suggested to be an activator for fruit set of zucchini squash (Martínez et al., 2013) and tomato (Shinozaki et al., 2015). Studies in orchids revealed that ethylene is also required for ovary development and ovule differentiation together with auxin (Zhang and O'Neill, 1993).
During fruit set, expression of most members of the ethylene synthesis genes, including ACC oxidase, was up-regulated in 2 daa fruitlets compared with 2 dba ovaries in WT and SlINVINH1-RNAi plants (Table 6). Consistently, levels of methionine, the precursor of ethylene, more than doubled in fruitlets compared with ovaries in both genotypes (Table 7), indicating positive roles played by ethylene for tomato fruit set. Interestingly, previous transcriptome studies found that mRNA levels of ethylene biosynthesis and signalling genes decreased in 4 daa fruitlets as compared with tomato ovaries at 2 dba (Wang et al., 2009) and at 0 dba (Pattison et al., 2015). These authors compared gene expression profiles of ovaries with 4 daa fruitlets, 2 d older than the 2 daa fruitlets used in our study. The discrepancy in the expression changes of ethylene synthesis and perception genes between our work and those reported by Wang et al. (2009) andPattison et al. (2015) likely comes from the different developmental stages at which the samples were harvested and analysed. Recently, ethylene production rate was measured directly by gas chromatography in pistils during tomato fruit set from 2 dba to 4 daa (Shinozaki et al., 2015). However, the pistil is the entire female organ, which could mask the molecular and biochemical status of the ovaries.
In most species, pollination is accompanied by an increase in ethylene production in the stigma and style hours after pollination and there is a burst of ethylene soon after fertilization (O'Neill, 1997;Lin et al., 2009). The up-regulated expression of ethylene synthesis genes during fruit set observed in our study might reflect a short ethylene burst following fertilization. On the other hand, since tomato petals and styles start to senescence soon after fertilization, an increased expression of genes encoding the ACC oxidase enzyme in ovaries ( Table 6) may increase ethylene levels, which could induce senescence of petals and styles (Llop-Tous et al., 2000).
In the 2 dba ovaries, we found that four of five genes encoding ACC oxidase increased their mRNA levels in CWINelevated plants ( Table 5). The up-regulated expression of these genes by increased CWIN activity may contribute to ovule and ovary development in tomato. Consistent with this proposition, silencing ACC oxidase inhibited tobacco ovule development (De Martinis and Mariani, 1999) and enhanced expression of ACC genes has been observed in ovaries during their developmental progression to anthesis in tomato (Vriezen et al., 2008).
It is also noteworthy that expression of two cyclin B2 genes were up-regulated in 2 dba ovaries of SlINVINH1-RNAi compared with WT ovaries (Table 5). CWIN-mediated release of hexose plays a role in cell division in seed and fruit development (Cheng and Chourey, 1999;Ruan, 2012Ruan, , 2014. The B-type cyclins are mitotic cyclins that mediate progression of cells into Fig. 6. A model of how CWIN-mediated sugar signalling may regulate fruit set. See Discussion for more details. Arrows indicate genes involved in the specified process that were up-regulated (red) or down-regulated (green) by elevated CWIN activity. Solid line indicates known pathway; dashed line indicates predicted pathways. and out of mitosis (Inzé and De Veylder, 2006). Overexpression of cycb2;2 in rice promoted root growth by increasing cell number (Lee et al., 2003). Cycb2;2 has been recently reported to function in establishing rice grain size and yield (Ishimaru et al., 2013) and cell wall formation in mitotic cells of maize endosperm (Sabelli et al., 2014). Thus, up-regulated expression of two cyclinB genes in tomato ovaries by elevated CWIN activity may stimulate the cell cycle during fruit set.

A model of CWIN mediating fruit set
Based on the above analyses, we propose a model for how CWIN may regulate fruit set (Fig. 6). Here the Glc and Fru produced from CWIN activity in the extracellular space could be (i) sensed by an unknown Glc sensor located on the plasma membrane (PM) to regulate gene expression and (ii) be taken up by the hexose transporter into the cytosol for metabolism or sugar signalling (Ruan et al., 1995Ruan 2014). Elevation of CWIN activity up-regulates expression of genes encoding proteins involved in defense (R genes), protein synthesis, cell cycle, and cell wall biosynthesis/remodelling, but down-regulates photosynthesis genes and receptor-like kinase (RLK) in ovaries of SlINVINH1-RNAi compared with WT plants. The genes encoding proteins functioning in cell wall biosynthesis/remodelling may cause changes in cell wall components, which may be sensed by RLK located on the plasma membrane to regulate gene expression through a small GTPase (ROP) pathway. Together, the model provides insights into the molecular pathways by which CWIN positively regulates fruit set.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Invertase activities in 2 dba ovaries and 2 daa fruitlets in the SlINVINH1-RNAi plants as compared with the WT. Fig. S2. Validation of RNA-seq data by qRT-PCR. Table S1. Primers used for qRT-PCR analyses. Table S2. A total of 5183 differentially expressed genes in wild type at fruit set Table S3. A total of 3262 differentially expressed genes during fruit set of SlINVINH1-RNAi plants. Table S4. A total of 497 differentially expressed genes in SlINVINH1-RNAi fruit set in response to elevated CWIN activity compared with wild type fruit set. Table S5. A total of 319 differentially expressed genes in ovaries of SlINVINH1-RNAi compared with those of wild type plants.