TET3‐mediated demethylation in tomato activates expression of a CETS gene that stimulates vegetative growth

Abstract Expression of the mammalian DNA demethylase enzyme TET3 in plants can be used to induce hypomethylation of DNA. In tomato lines that express a TET3 transgene, we observed distinct phenotypes including an increase in the length and number of leaves of primary shoots. As these changes resemble phenotypes observed in plants with strong expression of SELF PRUNING (SP), a member of the PEBP/CETS family, we investigated in TET3 lines the expression levels of members of the PEBP/CETS gene family, which affect shoot architecture and growth of sympodial units in tomato. We did not detect any changes in SP expression in TET3 lines, but for CEN1.1, a putative family member that has not been functionally characterized, we identified changes in gene expression that corresponded to hypomethylation in the upstream region. In tomato wild type, CEN1.1 is expressed in roots, petals, and shoot apices but not in mature leaves. In contrast, in TET3 transformants, the CEN1.1 gene became hypomethylated and activated in leaves. Ectopic expression of CEN1.1 in tomato caused similar phenotypes to those seen in TET3 transformants. Vegetative growth was increased, resulting both in a delay in inflorescence development and in an instability of the inflorescences, which frequently reverted to a vegetative state. Ectopic expression of CEN1.1 in Arabidopsis thaliana also caused floral repression. Our data suggest that the phenotypes observed in TET3 lines are a consequence of ectopic activation of CEN1.1, which promotes vegetative growth, and that CEN1.1 expression is sensitive to DNA methylation changes.

crop plants such as tomato, shoot architecture also has great economic importance, with different patterns being preferred for different purposes. For example, for mechanically harvested processing tomatoes, tomato plants with determinate growth have a higher yield, while tomato varieties that grow indeterminately are better suited to produce tomatoes that are eaten fresh and require continuous market delivery (Jiang et al., 2013). Tomato is an example of a plant species with a sympodial growth pattern, composed of a series of determinate meristems. The primary shoot of tomato terminates with an inflorescence after 8-12 compound leaves (McGarry & Ayre, 2012), but growth continues from the uppermost axillary meristem (Lifschitz et al., 2006). After this point, the shoot is formed from repeating sympodial units consisting of three leaves and terminating with an inflorescence. Upward growth of the shoot is again continued from the most proximal axillary bud of the previous sympodial unit in an indeterminate fashion (Lifschitz et al., 2006).
The establishment of this pattern relies on the balance between the expression levels of genes in the tomato PEBP gene family (phosphatidylethanolamine-binding protein), also called the CETS (CEN-TRORADIALIS/TERMINAL FLOWER 1/SELF PRUNING) gene family after its founding members (Shalit et al., 2009). This family is present in a large variety of species where it plays a role in mechanisms as diverse as bulb induction in onions and formation of needles in Norway spruce (Karlgren, Gyllenstrand, Clapham, & Lagercrantz, 2013;Lee, Baldwin, Kenel, McCallum, & Macknight, 2013;Wickland & Hanzawa, 2015). SFT (SINGLE FLOWER TRUSS), the tomato homolog of the Arabidopsis thaliana gene FT (FLOWERING LOCUS T; Lifschitz et al., 2006), and SP (SELF PRUNING), the tomato homolog of the Arabidopsis gene TFL1 (TERMINAL FLOWER 1; Pnueli et al., 1998), are the best described of the genes in this family in tomato. Mutations in SFT result in delayed flowering (Lifschitz et al., 2006) while sp tomato mutants initially flower after the normal number of leaves has been produced but afterward flowers switch to determinate growth (Shalit et al., 2009). Overexpression of the SFT gene causes early flowering, the opposite phenotype to overexpressing SP, which results in delayed termination of the primary shoot and increased numbers of leaves per sympodial unit (Lifschitz et al., 2006;McGarry & Ayre, 2012;Pnueli et al., 1998). Analysis of double mutants indicates that SP counteracts the florigenic effect of SFT in a dosageresponsive manner (Molinero-Rosales, Latorre, Jamilena, & Lozano, 2003;Shalit et al., 2009). In addition to SP and SFT, there are several other recognized members of the CETS gene family in tomato (Cao et al., 2016;Carmel-Goren, Liu, Lifschitz, & Zamir, 2003). Three of these (SP5G, SP5G2, and SP5G3) have been shown to have a role in delaying flowering, with knockdown lines of these genes showing early flowering and overexpression in Arabidopsis causing delayed flowering (Cao et al., 2016;Chitwood et al., 2013). Expression of SP5G, SP5G2, and SP5G3 is affected by day length (Cao et al., 2016).
Understanding the role of the genes in this family is an important tool to improve tomato crop yield or harvest index (yield per plant weight; Park et al., 2014;Soyk et al., 2017).
The likelihood of gene expression is frequently affected by epigenetic modifications to the gene, such as histone modifications and DNA methylation (Zilberman, Gehring, Tran, Ballinger, & Henikoff, 2007). DNA methylation occurs through the action of DNA methyltransferases and the presence of DNA methylation in the promoter of a gene is usually repressive, resulting in the silencing of that gene.
Given the importance of tomato as a crop plant and the involvement of methylation in the ripening process of tomato (Liu et al., 2015;Zhong et al., 2013), a better understanding of the role of methylation in tomato is extremely important. Expression of the catalytic domain of the mammalian DNA demethylase TET3 (TET3c) in Arabidopsis has previously been shown to be capable of causing DNA demethylation (Hollwey, Watson, & Meyer, 2016).
Here, by transforming the TET3c construct into tomato, we observed specific phenotypes and demonstrated that expression of CEN1.1, a member of the CETS gene family, is affected by DNA methylation upstream of the start codon. We show that hypomethylation caused by TET3c results in the activation of this CETS family member. We demonstrate that ectopic expression of either TET3c or CEN1.1 causes common phenotypes in tomato plants, including an instability of the transition to an inflorescence, delayed growth, and an increase in the number of leaves between inflorescences. Ectopic expression of CEN1.1 in Arabidopsis thaliana also results in an increase in the number of rosette leaves and a delay in flowering.

| Vector construction and plant transformation
The TET3c vector was constructed as described in Hollwey et al. (2016). The CEN1.1 vector was constructed by amplification of the CEN1.1 genomic region from tomato DNA using primers GGG AAGCTTGGCACGTTGATTGGTTTTTCG + GGGAATTCACAAGCAAA TGAGTAGGACAAACA. It was then cloned into the HindIII/EcoRI site of pGreen II 0029. The vectors were transferred into Agrobacterium tumefaciens for leaf disk transformation (Rai et al., 2012) of a EZCBT1 tomato variety and floral dip transformation of Arabidopsis thaliana (Col-0;Clough & Bent, 1998). Tomato transformation was carried out at the premises of ENZA ZADEN, Enkhuizen, The Netherlands.

| Plant material
Plants were grown in a growth chamber under long day conditions (16 hr light, 8 hr dark, 23°C, 42% humidity). At the age of 5 weeks, tomato plants were transferred to a glasshouse. All samplings for nucleic acid extractions were done between 8 and 10 a.m. to avoid possible circadian variations in gene expression or DNA methylation.

| Expression analyses
RNA for expression analysis was extracted as described in Stam et al. (2000). DNA was removed using the TURBO DNase kit (Ambion applied Biosystems) and converted to cDNA using M-MLV reverse transcriptase and oligo-dT primers (Invitrogen) according to the manufacturer's instructions. Semiquantitative PCR was carried out using MyTaq Red DNA Polymerase (Bioline) and qPCR was carried out using SsoFast Eva Green Supermix (Bio-Rad) according to the manufacturer's instructions. cDNA levels were normalized using eukaryotic translation initiation factor 3 primers GAGCGATGGAT GGTGAATCT + TTGTACGTGCGTCCAGAAAG.

| DNA methylation analysis
Genomic DNA for bisulfite sequencing was extracted according to Vejlupkova and Fowler (2003) with some modifications. Tissue for the SAP methylation analysis was isolated using a dissection microscope from FFPE sections of tomato shoot apices made according to Vitha, Balu ska, Jasik, Volkmann, and Barlow (2000). Bisulfite treatment was carried out using the EZ DNA Methylation-Lightning kit (Zymo Research). Bisulfite-treated DNA was amplified using primers AAYTTTTGGGGTGTGAGTTAGA + TCCACCCATTTCATTAACCACC and GTGAGGTGGGGTGTTAAAGAATGA + CACCRATRTAACACTC CACCT to amplify part of the region upstream of the CEN1.1 gene.
Oxidative bisulfite sequencing was performed as described in (Booth et al., 2013) to quantify levels of 5-methylcytosine and subtracted from bisulfite sequencing data (which contains 5-methylcytosine and 5-hydroxymethylcytosine) to calculate levels of 5-hydroxymethylcytosine; 10-20 clones were sequenced per sample. Sequencing data were analyzed using the online CYMATE tool (Hetzl, Foerster, Raidl, & Scheid, 2007)   To confirm that the reduction in 5mC levels was caused by TET3c, we screened the region for 5-hydroxymethylcytosine, a derivative of 5-methylcytosine produced by TET3 oxidation, which serves as a marker for TET3cmediated demethylation (Ito et al., 2010). Oxidative bisulfite sequencing showed that a significant increase in levels of 5-hydroxymethylcytosine occurred in TET3c tissue compared to wild-type tissue ( Fig. S2c).
We used semiquantitative RT-PCR to analyze the expression patterns of CEN1.1 in wild-type tomato. CEN1.1 was not expressed in plant leaves in both juvenile (5 weeks old) and mature (20 weeks old) tomato plants, but was expressed strongly in the shoot apex and also weakly in roots and petals (Figure 1c). Bisulfite sequencing was used to analyze whether expression of CEN1.1 correlated with hypomethylation in wild-type tissues as it does in TET3c plants.
Methylation levels were reduced in root and shoot apex (SAP) tissue where CEN1.1 is expressed, in comparison with leaf tissue where CEN1.1 is silenced (Figure 1d, Fig. S2d). In TET3c, root and SAP tissues, hypomethylation was observed at CHH sites, while overall CG and CHG methylation did not change significantly (Figure 1e).   Reduced methylation levels were seen in root and SAP DNA where CEN1.1 is expressed using the same 206-bp region previously analyzed in TET3c tomato. Bisulfite sequencing was used to analyze methylation levels in 5-week-old leaf, root, and shoot apex from wild-type tomato. Three biological replicates were averaged for each tissue. CG and CHG sites are indicated; all unlabeled sites are CHH sites. Points where methylation is significantly different are marked with a star (p < .05, calculated using Student's two-tailed t test). (e) Methylation levels were reduced in the CHH context in TET3c leaves, wild-type roots and wild-type SAP. Graphs show averages with error bars representing standard error. *p < .05, **p < .005, ***p < .0005, ns = not significant, calculated using Student's two-tailed t test meristem growing from the top of the fruit in 0.8% of fruit (n = 352; were leafy in comparison with control plants (0%, n = 17), a phenotype which had also been seen in 18% of TET3c inflorescences (n = 51; Fig. S1d). Inflorescences were classified as leafy when they contained multiple leaves and at least one vegetative meristem. Inflorescences containing leaves have also been described for lines that overexpress SP, although the reported effects are less severe than the ones we observed (Pnueli et al., 1998), and in sft, macrocalyx, or jointless mutants (Quinet, 2006;Vrebalov et al., 2002). Expression of these genes remains unchanged in the 35S::CEN1.1 tomato, which argues against CEN1.1 overexpression altering their expression (Fig. S3).
While these abnormal, leafy inflorescences contained large quantities of vegetative material, they also produced a larger number of flowers due to the large branched nature of the inflorescence. Therefore, 35S:: CEN1.1 inflorescences also produce more flowers on average than wild-type inflorescences (Figure 4a), although the number of flowers on an inflorescence varied significantly, ranging from 11 to 60. This phenotype becomes more obvious when vegetative material is removed during the development of the inflorescence (Figure 4b).
While it required more time for 35S with some plants being delayed for over 10 months, and one never flowering at all (Amaya, Ratcliffe, & Bradley, 1999). The delay in flowering caused by CEN1.1 in Arabidopsis was more severe than was observed when the Arabidopsis CEN1.1 homologues, TFL1 and BFT, were expressed under the 35S promoter (Mimida et al., 2001;Yoo et al., 2010). As would be expected, ectopic expression of the tomato SFT gene in Arabidopsis has the opposite effect to CEN1.1, causing early flowering after the production of four rosette leaves (Cao et al., 2016). with several proteins in tomato including a kinase, 14-3-3 proteins and a putative bZIP transcription factor (Pnueli et al., 2001; as does FT in Arabidopsis (Abe et al., 2005)), but the full pathway has not been elucidated and it is not known how the other known floral repressors in the tomato CETS gene family exert their influence (Cao et al., 2016). The development of leaves on inflorescences induced by ectopic expression of CEN1.1 had also been observed in sft, macrocalyx, and jointless mutants (Quinet, 2006;Vrebalov et al., 2002). There was no indication that any of these genes altered their expression in CEN1.1 transformants, which argues against their involvement in the phenotype observed in the transformants.

| Increased vegetative growth caused by
CEN1.1 appears differently in various tissues and paradoxically increases total fruit yield Ectopic expression of CEN1.1 also stimulates vegetative growth elsewhere in tomato plants, which seems to be more severe than similar phenotypes described in 35S::SP tomato plants (Quinet, 2006  although often at a lower frequency or intensity. This is to be expected, given that CEN1.1 expression due to TET3c-mediated demethylation is likely to be less intense than the strong, constitutive expression under the 35S promoter. Identification of the CEN1.1 gene illustrates that TET3c expression is a useful tool to discover previously unknown plant genes that are affected by DNA methylation changes. These may be otherwise difficult to detect, especially in species which are particularly susceptible to changes in DNA methylation. Arabidopsis mutants of the main methyltransferases are still viable. This allows high-throughput analysis of changes in DNA methylation and gene expression, which can identify genes controlled by methylation. In contrast, species such as tomato and rice appear to be more sensitive to DNA methylation changes as they show more adverse effects when the enzymes involved in DNA methylation are lost (Liu et al., 2015;Ono et al., 2012). In tomato, null mutations of SlNRPE1, a component of the RdDM pathway, are lethal (Gouil & Baulcombe, 2016), and MET1 RNAi lines are not viable (Watson, 2013), making the identification of genes and processes affected by DNA methylation changes more challenging. TET3c expression may therefore offer an alternative in