DNA Methylation Involved in Pepper (Capsicum annuum L.) Fruit Ripening Regulation and Interacts with Phytohormones.

For understanding the regulatory mechanism of pepper fruit ripening and clarifying whether this process involves DNA methylation, experiments at the transcriptional, epigenetic and physiological levels were conducted. McrBC-PCR, bisulfite sequencing and real-time PCR demonstrated that DNA hypomethylation occurred in the UROT of some pepper ripening-related genes at the turning stage, which may be attributed to the up-regulation of CaDML2-like and the down-regulation of CaMET1-like1, CaMET1-like2, CaCMT2-like and CaCMT4-like. Silencing of CaMET1-like1 by VIGS led to DNA hypomethylation, increased SSC and accumulation of carotenoids in pepper fruit, which accompanied with the expression changes of genes involved in capsanthin-capsorubin biosynthesis, cell wall degradation and phytohormone metabolism and signaling. Endogenous ABA increased during pepper fruit ripening, whereas endogenous IAA showed an opposite trend. No ethylene signal was detected during pepper ripening. DNA hypomethylation repressed the expression of auxin and gibberellin biosynthesis genes as well as the cytokinin degradation genes, but induced the expression of ABA biosynthesis genes. In mature-green pericarp, exogenous ABA induced the expression of CaDML2-like, but repressed that of CaCMT4-like. IAA treatment promoted the transcription of CaMET1-like1 and CaCMT3-like. Ethephon significantly up-regulated the expression of CaDML2-like. GA3 and 6-BA treatments showed indistinct effects on DNA methylation at the transcriptional level.


Introduction
The ripening of fleshy fruit involves an elaborate process that is regulated by many factors at multiple levels. There is growing evidence to suggest that epigenetic tags are a critical regulator of fruit ripening, especially DNA methylation (Zhong et al., 2013;Liu et al., 2015;Lang et al., 2017;Cheng et al., 2018). DNA methylation mainly refers to the methyl at the 5´ position of cytosine (5mC) (Bender, 2004). Based on the neighboring nucleotides, it can be classified into symmetric (CG or CHG) and asymmetric (CHH) types (Deleris et al., 2016). The formation of asymmetric DNA methylation is catalysed by CHROMOMETHYLASE 2 (CMT2) and DOMAIN REARRANGED METHYLTRANSFERASES (DRM1 and DRM2) through the RNA-directed DNA methylation (RdDM) pathway, whereas the symmetric type is catalysed by METHYLTRANSFERASE 1 (MET1) and CHROMOMETHYLASE 3 (CMT3) (Finnegan and Kovac, 2000;Lindroth et al., 2001;Bender, 2004;Zhang et al., 2010;Deleris et al., 2016). DNA methylation can be passively lost after DNA replication in the absence of methyltransferase activity or it can be actively eliminated by the DNA demethylases REPRESSOR OF SILENCING 1 (ROS1), DEMETER (DME), DME-LIKE 2 (DML2), and DME-LIKE 3 (DML3), which remove the 5mC base and then fill the single-nucleotide gap with a non-methylated cytosine (Gong et al., 2002;Hsieh et al., 2009;Wu and Zhang, 2010;Liu et al., 2015;Lang et al., 2017).
DNA methylation has been demonstrated to regulate fruit ripening in tomato, strawberry, and sweet orange. A global decrease of DNA methylation, especially at the 5´-and 3´-regulatory regions of the genes, has been associated with tomato and strawberry fruit ripening, and artificially accelerating this process by application of the methyltransferase inhibitor 5-azacytidine induces premature ripening (Zhong et al., 2013;Cheng et al., 2018). In contrast, inhibition of ripening-associated hypomethylation through knockdown of SlDML2 disturbs the natural dynamic expression patterns of many tomato genes related to ripening and ultimately leads to retarded ripening Lang et al., 2017). In the case of sweet orange, a global hypermethylation of DNA is found in the peel, which is critical for the ripening-associated process of de-greening (Huang et al., 2019). Differentially expressed genes regulated by ripening-associated changes in DNA methylation are usually enriched in the processes of pigment biosynthesis (such as anthocyanin biosynthesis in strawberry and carotenoid biosynthesis in tomato), aromatic compound biosynthesis, cell wall degradation, and ripeningrelated phytohormone biosynthesis (such as ABA biosynthesis in strawberry and sweet orange, and ethylene biosynthesis in tomato) (Lang et al., 2017;Cheng et al., 2018;Huang et al., 2019). The ripening-associated variation in DNA methylation is controlled by DNA methyltransferase and demethylase genes at the transcriptional level. In tomato, the ripeningassociated DNA hypomethylation is attributed to both the down-regulation of DNA methyltransferase genes and the up-regulation of DNA demethylase genes (Cao et al., 2014). On the other hand, in strawberry this epigenetic change is mainly caused by the down-regulated of genes involving in RdDM (Cheng et al., 2018). In the peel of sweet orange, ripening-associated DNA hypermethylation is primarily caused by the down-regulation of DNA demethylase genes (Huang et al., 2019).
In contrast with the recent evidence for the involvement of epigenetic tags, the role of phytohormones in fruit ripening is long established. Ethylene plays a critical role in triggering ripening in climacteric fruit, and suppression of its biosynthesis and/or blocking of its signal pathway severely delays the normal ripening process (Oeller et al., 1991;Lanahan et al., 1994;Li et al., 2007;Lee et al., 2012;Sekeli et al., 2014). ABA is well accepted as an important positive regulator of ripening in both climacteric and non-climacteric fruit Sun et al., 2012aSun et al., , 2012bJi et al., 2014;Wang et al., 2013Wang et al., , 2017. The accumulation of ABA is associated with ripening in many fleshy fruits, such as tomato (Sun et al., 2012a(Sun et al., , 2012b, cucumber (Wang et al., 2013), watermelon , grape (Sun et al., 2010), and strawberry . Artificially inhibiting ABA accumulation and/ or blocking its perception results in retarded fruit ripening Jia et al., 2011;Sun et al., 2012aSun et al., , 2012bJi et al., 2014). Auxin has been found to have contrasting functions in the regulation fruit ripening. In most cases, it inhibits ripening by antagonizing the effects of ethylene or ABA (Manning et al., 1994;Davies et al., 1997;Liu et al., 2005;Böttcher et al., 2011;Symons et al., 2012;Ziliotto et al., 2012;Hao et al., 2015;Su et al., 2015;Li et al., 2016Li et al., , 2017Liao et al., 2018;Sravankumar et al., 2018); however, in some cases, especially in climacteric fruits such as tomato, peach, and fig, auxin promotes ripening by inducing the expression of ethylene biosynthesis genes (Gillaspy et al., 1993;Jones et al., 2002;Owino et al., 2006;Trainotti et al., 2007). Gibberellin generally plays a repressive role in the regulation of fleshy fruit ripening through antagonizing the effects of ethylene (Dostal and Leopold, 1967;Biale, 1978;Martínez et al., 1994;Martínez-Romero et al., 2000;Alós et al., 2006;Kader, 2008). Our knowledge about the role of cytokinin in the regulation of fleshy fruit ripening is very limited, although it has been reported to repress ripening in grape berries (Peppi and Fidelibus, 2008).
Pepper (Capsicum annuum) is a fruit that is consumed worldwide and it is sold at the mature and ripening stages. Ripening determines pepper fruit quality because, on one hand, it promotes the accumulation of pigments and other nutrients that increase the quality of the ripe fruit, and on the other hand, post-harvest ripening and coloration severely reduce fruit quality and shelf-life of mature-green peppers. To date, our understanding of pepper fruit ripening is very limited. Pepper is currently considered as a medium type between climacteric and non-climacteric fruits (Villavicencio et al., 2001) since some varieties, especially hot peppers, show an ethylene burst during ripening (Villavicencio et al., 1999;Hou et al., 2018) whereas some varieties, such as sweet peppers, do not (Lurie and Ben-Yehoshua 1986;Pretel et al., 1995;Watkins, 2002). The regulatory mechanism of pepper fruit ripening remains unclear, although a few studies have found that ethylene and ABA seem to promote ripening whilst auxin, gibberellin, and cytokinin are likely to play opposite roles (Deruère et al., 1994;Hou et al., 2018). Whether DNA methylation affects the regulation of ripening and interacts with phytohormones in pepper is even more unclear. In this study, we therefore examined the effects of DNA methylation on pepper fruit ripening and investigated its interactions with phytohormones. Our results demonstrated that the degree of DNA methylation in the fruit may be controlled by DNA methyltransferase and demethylase genes at the transcriptional level. DNA hypomethylation caused by silencing of CaMET1-like1 led to premature ripening. DNA methylation interacted with phytohormones in a way that may fine-tune the ripening of the fruit.

Plant material
The sweet pepper (Capsicum annuum L.) inbred line 16C391 was developed in our lab and bears non-climacteric fruit. Plants used for developmental studies were grown in a greenhouse under natural daylight at the Experimental Station of China Agricultural University in 2017 and 2018. Plants were supplied with adequate water and nutrients according to standard horticultural practice. Pericarps were collected at the following stages: immature-green (IM, 15 d post anthesis, DPA), maturegreen (MG, 38 DPA), breaker (B, 44 DPA), turning (T, 50 DPA), and red-ripening (R, 55 DPA). They were immediately frozen with liquid nitrogen and stored at -80 °C.

Bisulfite sequencing
Bisulfite sequencing was performed to further confirm the DNA methylation level. DNA bisulfite conversion was conducted with 1 µg of genomic DNA using the a DNA Bisulfite Conversion Kit (Tiangen, China) following the manufacturer's instructions. Methylation-specific PCR was set up with ~100 ng converted or non-converted genomic DNA as the template using a Methylation-specific PCR Kit (Tiangen, China) according to the user's manual. Primers for the methylation-specific PCR were designed within the McrBC-PCR tested regions and are listed in Supplementary Table S2. The length of the amplification fragments ranged from 150-300 bp ( Supplementary Fig. S1). Since every cytosine can be potentially methylated, to avoid any sequence selection bias during the PCR, the 'C' and 'G' nucleotides were replaced by 'Y' and 'R' in the forward and reverse primers, respectively. Ten single colonies for each PCR fragment were sequenced. The methylation ratio for each C-G site was calculated by dividing the number of non-changed nucleotides by the sequencing depth.

Quantitative real-time PCR analysis
Total RNA extraction, first strand cDNA synthesis, and real-time PCR were conducted according to Sun et al. (2011). UBI-3 (Capana06g002873) was selected as the internal control (Wan et al., 2011) and relative expression values were calculated using the 2 -∆∆CT method (Livak and Schmittgen, 2001). For each gene, three biological replicates were performed, and for each biological replicate at least three technical replicates were performed. Primers used in this analysis are listed in Supplementary  Table S3.  The coding sequence fragments of CaMET1-like1, CaCMT3-like, and CaDML2-like without the stop codon were PCR-amplified and subcloned into the Super1300 vector in frame with green fluorescence protein (GFP) driven by the CaMV 35S promoter. Plasmids were transferred into onion epidermal cells using particle bombardment according to Lee et al. (2006). GFP fluorescence was detected and captured using a Carl Zeiss LSM 510 system. Each assay was repeated three times. The primers used are listed in Supplementary Table S4.

Determination of IAA, ABA, and ethylene levels
Levels of auxin (IAA) and ABA were determined in the pericarp of normally developing fruits at the IM, M, T, and R stages and also in premature-ripe and green pericarps of the CaMET1-like1-silenced fruit using UPLC-MS (Waters Corporation and ThermoFisher Scientific) according to Wang et al. (2017). Ethylene levels of normally developing fruit were measured at the IM, M, T, and R stages using gas chromatography (GC 17A, Shimadzu) according to Xue et al., (2008).

Virus-induced gene silencing
A VIGS system based on the tobacco rattle virus (TRV) was used to investigate gene function (Liu et al., 2002). Fragments of CaMET1-like1 (423 bp) and CaPDS (452 bp) were PCR-amplified from pepper cDNA and cloned into the pTRV2 vector to generate the plasmids pTRV2-CaMET1-like1 and pTRV2-CaPDS, respectively. pTRV1, pTRV2, pTRV2-CaPDS, and pTRV2-CaMET1-like1 were introduced into Agrobacterium tumefaciens strain GV3101 by electroporation. The transformed cells were grown for 8-10 h at 28 °C in Luria-Bertani (LB) medium containing the appropriate antibiotics, and were then collected by centrifugation (5000 g for 15 min), re-suspended in infiltration buffer (10 mM MgCl 2 , 10 mM MES, and 100 µM acetosyringone), and incubated at 28 °C for 3 h. pTRV1 and pTRV2 or its derivatives were mixed at a 1:1 (v/v) ratio and infiltrated into cotyledons of 1-week-old pepper seedlings using a 1-cm 3 syringe without a needle (Liu et al. 2002). The infiltrated plants were grown in a culture room at 22 °C, 16/8 h light/dark, 150 μmol m -2 s -1 . Since silencing of CaPDS leads to light bleaching of green organs, plants infiltrated with pTRV1+pTRV2-CaPDS were used as the positive control to check the activity of pTRV1 and to indicate the spatio-temporal pattern of gene silencing (Guo et al., 2018). When light-bleaching appeared in the positive control, reverse-transcription PCR was conducted to detect pTRV1, pTRV2 and its derivatives in all the infiltrated plants. Plants carrying pTRV1+pTRV2-CaMET1-like1 and pTRV1+pTRV2 (negative control) were used for further analysis. Flowers at anthesis were hand-pollinated and tagged, and the fruit ripening time was evaluated as the number of days from anthesis to the breaker stage. At least 100 plants were infiltrated with pTRV1+pTRV2-CaMET1-like1 and at least 30 plants were injected with negative or positive control plasmid mixtures. The whole experiment was repeated twice. The primers used are listed in Supplementary Table S5.

RNA-seq analysis
Premature-ripe and green pericarps of the CaMET1-like1-silenced fruit and red-ripe pericarps of the negative control were collected from the two biological repeats as well as from a preliminary experiment. RNA extraction and library construction were conducted according to Zhong et al. (2011). Sequencing was conducted by the Biomarker Technologies Corporation (Beijing, China) on a GAII platform (Illumina) and pairedend reads were generated. Filtration of the reads, alignment, and expression calculations were carried out according to Huang et al. (2013). Differentially expressed genes (DEGs) were calculated using DESeq2 in R and were then screened according the following conditions: the average expression value of each gene was greater than 2 fragments per kilobase of transcript per million fragments mapped reads in at least one of the three samples, and the expected false positive rate value was less than 0.05. The KEGG statistics of pathway enrichment were analysed and plotted using BMKCloud (http://www.biocloud.net/) with the default settings. Principal component analysis (PCA) was conducted with expression profiles of all the samples using the stats::prcomp() function in R. Prior to that, data were mean centered and Z-scored. The PCA results were plotted and edited using ggplot2::ggplot() (Wickham, 2016) in R and Photoshop CS5, respectively.

Ripening-related parameters
Soluble solids content (SSC) and titratable acidity (TA) were measured using 50-100 g pericarp tissue taken from 3-5 fruit according to Kong et al. (2013). Carotenoids were extracted from 50 mg freeze-dried pericarp samples and measured using aLC-APCI-MS/MS system (AB SCIEX, USA) according to Giuffrida et al. (2013). Standard substances examined in study are listed in Supplementary Table S6. Three or more replicates of each assay were performed.

Phytohormone treatments
Fruit were harvested at the MG stage, surface-sterilized with 2.5% (w/v) NaClO aqueous solution for 15 min, and then washed three times with sterile water. Pericarp discs (radius 0.6 cm) were cut from the equatorial region of the fruit using a hole puncher. The discs were soaked in MES buffer (pH=5.5) containing 5 mmol l -1 CaCl 2 , 5 mmol/l -1 MgCl 2 , 5 mmol/l -1 EDTA, 5 mmol/l -1 vitamin C, and 200 mmol/l -1 mannitol for 30 mins (Télef et al., 2006). Phytohormones were then added into the buffer to give final concentrations of 25 μM for auxin (IAA; Sigma-Aldrich I5148), 100 μM for ABA (Sigma-Aldrich A1049), 50 μM for gibberellic acid (GA 3 ; Sigma-Aldrich G7645), 25 μM for cytokinin (6-BA; Sigma-Aldrich B3408), and 25 μM for ethephon (Sigma-Aldrich 45473). The optimal concentration for each hormone was determined in preliminary experiments. At 4 h after adding the hormones the pericarp discs were collected, frozen with liquid nitrogen, and stored at -80 °C. For each treatment, at least 15 discs were used and the whole experiment was repeated three times. Expression of the DNA methyltransferase and demethylases genes, ripening-related genes (CaPSY1, CaPDS, CaPG, CaNCED1, and CaCYP707A1) together with several hormoneresponsive genes (CaARF2 for IAA, CaABI2 for ABA, CaDELLA for GA 3 , CaARR5 for 6-BA, and CaETR1 for ethylene) were analysed using real-time PCR. Primers used in this experiment are listed in Supplementary Table S3.

Statistical analyses
Significant differences between two groups of samples were determined using Student's t-test via the t.test function in Excel. Significant differences among multiple groups of samples were tested by ANOVA and post hoc Tukey's HSD test using stats::anova() and agricolae::HSD.test(), respectively, in R (https://CRAN.R-project.org/package=agricolae). For real-time PCR, a significant difference between two groups of samples had to satisfy the following conditions: statistically significant difference at the 0.05 level, and the log 2 -transformed expression fold-change had to be either greater than 1 or less than 0.5.

DNA hypomethylation occurs in the UROT of several ripening-related genes during fruit ripening
To explore the link between DNA methylation and pepper fruit ripening, the DNA methylation levels of the upstream region of the transcriptional start site (UROT) of several ripeningrelated genes were determined in the pericarp at the IM and T stages, namely those involved in capsanthin/capsorubin biosynthesis (CaPSY1, CaPDS, and CaCCS), cell wall degradation (CaPG,CaPME2.2,CaCEL1,and CaCEL2), ABA metabolism (CaNCED1 and CaCYP707A1) together with genes encoding ripening-related transcription factors (CaRIN, CaCNR, and CaNOR) (Fig. 1A, B). The expression of CaPSY1, CaPDS, CaCCS, CaPG, CaCEL1, CaCEL2, CaRIN, and CaNCED1 was up-regulated during ripening, whereas the expression of the other genes was down-regulated ( Fig. 1C, D). According to the results of the McrBC-PCR, DNA hypomethylation was most likely to occur in the UROT of CaPSY1, CaPG,CaPME2.2,CaRIN,CaCNR, and CaNCED1 at the T stage (Fig. 1A). In contrast, low levels of DNA methylation were detected in the UROT of CaPDS and CaCCS at the IM and T stages, whilst consistently high levels of DNA methylation were recorded in the tested regions of CaCEL1, CaCEL2, CaNOR, and CaCYP707A1 (Fig. 1A). Although the bisulfitesequencing results were not completely consistent with those of McrBC-PCR, various degrees (~12.42-61.84%) of DNA hypomethylation were recorded in the UROT of CaPSY1, CaCCS,CaPG,CaPME2.2,CaRIN,CaCNR,CaNCED1, and CaCYP707A1 at the T stage (Fig. 1B).

Identification and expression analysis of DNA methyltransferase and demethylase genes
Orthologs of Arabidopsis and tomato DNA methyltransferase and demethylase genes were identified in the pepper genome. Based on the phylogenetic tree ( Fig. 2A), they were renamed as . Motif analysis showed that CaCMTs, CaMETs, CaDRMs, and CaDMLs shared conserved functional domains with their Arabidopsis orthologs ( Fig. 2B; Supplementary Tables S7, S8), suggesting that they may play similar roles. CaMET1-like1, CaCMT3-like, and CaDML2-like were all localized in the nucleus, which further suggested their potential roles in the regulation of DNA methylation ( Supplementary Fig. S2). Expression analysis indicated that both CaMET1-like1 and CaMET1-like2 followed a similar pattern of decreased expression after the MG stage; however, at most stages CaMET1-like1 was expressed at a higher level than CaMET1-like2 (Fig. 2C). In the CaCMT family, expression of CaCMT2-like dramatically decreased from the IM to B stage, whereas expression of CaCMT4-like peaked at the MG stage and then decreased (Fig. 2D). The expression of CaCMT3-like was relatively low during the whole ripening period (Fig. 2D). In the CaDRM family, expression of CaDRM5-like1, CaDRM5-like2, and CaDRM8-like was significantly higher than that of CaDRM6-like at most stages (Fig. 2E). In particular, CaDRM5-like1 and CaDRM5-like2 were highly expressed at the MG and B stages, while expression of CaDRM8-like peaked at the T stage (Fig. 2E). In the CaDML family, expression of CaDML2-like, which peaked at the T stage, was significantly higher than that of the other members at most stages (Fig. 2F). The expression of CaDML1like, CaDML3-like, and CaDML4-like peaked at the IM, MG, and B stages, respectively. It can be assumed that CaMET1-like1, and CaDML2-like may play important roles in the regulation of DNA methylation during pepper fruit development and ripening.

Silencing of CaMET1-like1 promotes fruit ripening and leads to DNA hypomethylation
In order to further explore the effect of DNA methylation on ripening, CaMET1-like1 was silenced using VIGS. In replicates I and II, accelerated fruit ripening was observed in 21 and 20 plants, respectively, which was 4-8 d earlier than that of the negative control (Fig. 3A, B). The color transition in each precocious fruit was restricted to certain areas of the pericarp for several days and led to heterogeneous fruit color (Fig. 3C, D), which was different from that of the untreated fruit (Fig. 1D) and the negative controls (Fig. 3C). Heterogenous coloration was also observed in the CaPDS-silenced fruit (Fig. 3E). Later, the premature-ripe and green pericarps were sampled from the precocious fruits produced by the CaMET1-like1silenced plants V11, V49, and V72. RT-PCR indicated that the premature-ripe pericarp simultaneously carried the pTRV1 and pTRV2-CaMET1-like1 plasmids, whereas the green pericarp only harbored one of them (Fig. 3F). Consistent with this, expression of CaMET1-like1 in the premature-ripe pericarp decreased to less than 40% of that of the green pericarp (Fig. 3G). Besides the fruit, reduced CaMET1-like1 expression was also observed in the leaves of the three tested plants (Fig. 3H). Bisulfite sequencing demonstrated that the DNA methylation levels of the UROT of CaPSY1, CaCEL1, and CaNOR were down-regulated in most premature-ripe pericarps ( Fig. 3I-L). In accordance with the molecular data, premature-ripe pericarps had a higher level of soluble solids content (SSC) and accumulated more phytoene, β-carotene, zeaxanthin, antheraxanthin, capsanthin, capsorubin, and α-carotene than green pericarps (Tables 1, 2). However, when compared with red-ripe pericarps of the negative control, premature-ripe pericarps accumulated higher levels of phytoene, β-carotene, zeaxanthin, antheraxanthin, violaxanthin, α-carotene, and lutein (Table 2). In contrast, the SSC and carotenoid levels of green pericarps in silenced fruit were similar to those in mature-green fruit of the negative control (Tables 1, 2). With regards to phytohormones, premature-ripe pericarps accumulated more ABA and less IAA than green pericarps (Supplementary Table S9). It can therefore be concluded  . 'pTRV1/2M-RA' indicates injected plants that carried both the pTRV1 and pTRV2-CaMET1-like1 plasmids and which showed accelerated fruit ripening. 'pTRV1/2M' indicates injected plants that carried both the pTRV1 and pTRV2-CaMET1-like1 plasmids and which showed no effect on fruit ripening. 'ND' indicates injected plants that only carried one of the above plasmids. (B) Boxplot of the time of onset of ripening of fruit injected with pTRV1 + pTRV2-CaMET1-like1 (pTRV2M) and pTRV1 + pTRV2. Data are shown from two replicates (I, II). Student's t-test was used to determine the P-values. (C-E) Plants injected with pTRV1 + pTRV2 (C), pTRV1 + pTRV2-CaMET1-like1 (D), and pTRV1 + pTRV2-CaPDS (E). The white arrowhead in (D) indicates the coloration on a premature-ripe fruit. The yellow and white arrows in (E) indicate a light-bleached peduncle and pericarp, respectively. Light-bleached leaves are also shown in (E). (F) PCR-based detection of pTRV1 and pTRV2-CaMET1-like1 plasmids in three heterogeneously colored precocious fruits, which were collected from three replicate injected plants (V11, V49, and V72). 'R' and 'G' indicate prematureripe and green pericarps, respectively, collected from the same precocious fruit. (G) Relative expression of CaMET1-like1 in the premature-ripe (red bars) and green pericarps (green bars). Significant differences were determined using Student's t-test: **P<0.01. (H) Relative expression of CaMET1-like1 in leaves collected from three replicate plants injected with pTRV1 + pTRV2 and pTRV1 + pTRV2-CaMET1-like1 (VIGS11, VIGS49, and VIGS72). Significant differences were determined using ANOVA followed by Tukey's HSD test (P<0.01). that silencing of CaMET1-like1 resulted in premature ripening and DNA hypomethylation in the pepper fruit.

Silencing of CaMET1-like1 affects gene expression in the fruit
Comparative transcriptional analysis was conducted in order to investigate the effects of CaMET1-like1 on pepper fruit ripening. PCA showed that samples collected from red-ripe pericarps of the negative control and from premature-ripe and green pericarps of the silenced fruit were clearly separated in the plot defined by PC1 and PC2, which together explained ~80% of the variation ( Supplementary Fig. S3), suggesting that gene silencing affected the pepper fruit transcriptome. DEGs between the premature-ripe and green pericarp were mainly enriched in the categories of plant hormone signal transduction, pyruvate metabolism, arginine and proline metabolism, the TCA cycle, and steroid and cutin biosynthesis (Fig. 4A).DEGs between the premature-ripe pericarp and the red-ripe negative control pericarp were mainly enriched in phenylpropanoid biosynthesis, photosynthesis, DNA replication, porphyrin, and chlorophyll metabolism (Supplementary Fig. S4). Putative ripening-related DEGs were extracted from the two comparisons, and included those involved in DNA methylation, carotenoid biosynthesis, cell wall degradation, phytohormone (ABA, ethylene, auxin, GA, and cytokinin) metabolism and signaling, auxin polar transport, transcriptional regulation, and capsaicin biosynthesis (Fig. 4B, C). In particular, compared with the genes in the green pericarp of the silenced fruit, in premature-ripe pericarps CaMET1-like1 was down-regulated while CaDRM6like was up-regulated; CaPSY1 and CaCCS, which are involved in capsanthin/capsorubin biosynthesis, were up-regulated while Caε-LCY and Caβ-LCY were down-regulated; and CaPG, CaEXPA1, and CaCEL1, which are involved in cell wall degradation, were up-regulated. With regards to ethylene biosynthesis genes, CaACS10 and CaACO1 were down-regulated while CaACO3 was up-regulated. For auxin-related genes, most of the CaLAXs, CaIAAs, CaAUX, and CaARFs were down-regulated, except for CaGH3.1, CaARF2, CaARF5-1, and CaARF5-2, which were up-regulated; CaKAO1 and CaGAOX,1, which are involved in GA metabolism, were down-regulated. For capsaicin biosynthesis genes, expression of CaCOMT and CaBCAT2 was increased, while expression of Ca4CL1/2, CaHCT, and CaLACS8 was decreased. CaARR3/18, CaCRE1, CaCKH, and CaCKX5/6, which All data are μg g -1 fresh weight. ND, not detected.

Phytohormones affect the expression of genes related to DNA methylation and ripening in the pericarp
In order to investigate the relationships between phytohormones and genes related to DNA methylation during pepper fruit ripening, phytohormones were applied exogenously to discs of mature-green pericarp. Exogenous IAA clearly up-regulated the expression of CaMET1-like1 and CaCMT3-like, but down-regulated the expression of CaPG and CaNCED1 (Fig. 5A). Exogenous ABA repressed the expression of CaCMT4-like, but induced the accumulation of transcripts of CaDML2-like, CaPSY1, CaPG, CaNCED1, and CaCYP707A1. Gibberellin down-regulated the expression of CaDRM8-like, CaDML1-like, and CaDML3-like, but up-regulated the expression of CaCMT3-like. Application of 6-BA, a synthetic cytokinin, resulted in up-regulation of CaCMT2-like, CaDML2-like, and CaPG, and in down-regulation of CaMET1-like1. Ethephon clearly induced the expression of CaDML2-like, but repressed the expression of CaCYP707A1.
Dynamic changes of IAA, ABA, and ethylene levels during ripening In order to examine the effects of IAA and ABA on pepper fruit ripening, their concentrations were measured at different developmental stages. A decrease in endogenous IAA was observed from the IM to T stage (Fig. 5B). In contrast, the levels of endogenous ABA decreased slightly from the IM to MG stage and then increased and peaked at the R stage.
No ethylene signal was detected at any of the tested stages ( Supplementary Fig. S5).

DNA methylation may be involved in the regulation of ripening in pepper fruit
DNA methylation has been recently been discovered to regulate fruit ripening in tomato, strawberry, and sweet orange Lang et al., 2017;Cheng et al., 2018;Huang et al., 2019). In our present study, DNA hypomethylation was observed in the UROT of some ripening-related genes during pepper fruit ripening (Fig. 1A, B). Consistent with this, dynamic changes in the expression of DNA methyltransferase and demethylase genes associated with ripening were also observed ( Fig. 2C-F), suggesting that there was very likely a link between DNA methylation and fruit ripening. Moreover, silencing of CaMET1-like1 using VIGS resulted in DNA hypomethylation (Fig. 3I-L), premature accumulation of carotenoids (Table 2), increased levels of SSCs and ABA (Table 1, Supplementary  Table S9), and decreased levels of IAA (Supplementary Table  S9). The gene expression profile of premature-ripe pericarps of the CaMET1-like1-silenced fruit was also shifted towards to that of the red-ripe pericarps of the negative control ( Supplementary Fig. S3). In particular, capsanthin/capsorubin biosynthesis (up-regulation of CaPSY1 and CaCCS), cell wall degradation (up-regulation of CaPG, CaEXPA1, and CaCEL1; down-regulation of CaEXP13, CaPME22/61, and CaEG25), and ABA biosynthesis (up-regulation of CaNCED1) were promoted, whereas IAA signaling (downregulation of CaIAA8/9/13/16/27/29 and CaAUX22B/ D1/D2) and gibberellin biosynthesis (down-regulation of CaKAO and CaGAOX1) were repressed (Fig. 4B). Therefore, it can be deduced that CaMET1-like1, which is involved in the regulation of DNA methylation and hypomethylation in pepper fruit, may promote ripening. In accordance with this, ripening-associated DNA hypomethylation is also observed in tomato and strawberry, and its disturbance leads to abnormal fruit ripening Lang et al., 2017;Cheng et al., 2018). This suggests that the regulation of DNA methylation in fleshy fruit ripening may be common in some species. DNA hypomethylation occurs in the putative promoter region of PSY1 in tomato and pepper and it enhances the expression of the gene Lang et al., 2017;Fig. 1A, B), suggesting that PSY1 may be a key point in the regulation of fruit ripening by DNA methylation. The comparison between premature-ripe and redripe pericarps was also informative. Compared with redripe pericarps, the premature-ripe ones accumulated more phytoene, β-carotene, zeaxanthin, antheraxanthin, violaxanthin, α-carotene, and lutein (Table 2), whereas the levels of capsanthin and capsorubin levels were not changed significantly. A possible reason for this was the higher expression of CaDXPS, CaPSY2, and Caε-LCY in premature-ripe pericarps. The up-regulation of the first two genes may have led to the increase of total carotenoids, while the increased expression of Caε-LCY may have enhanced the δ-carotene-α-carotene-lutein pathway and ultimately led to high levels of α-carotene and lutein. However, since the expression of CaCCS and CaVDE was not included in the list of DEGs (Fig. 4B), the capsanthin and capsorubin levels may not have been significantly changed.

Transcriptional regulation of DNA methylation during ripening
The DNA methylation level is controlled by DNA methyltransferases (MET1, CMT2, CMT3, and DRMs) and demethylases (ROS1, DME, and DMLs) (Deleris et al., 2016;Gallusci et al., 2016). In our study, members of these families were identified in the pepper genome ( Fig. 2A), which is consistent with the hypothesis that DNA methylation is conserved in all flowering plants (Goll and Bestor, 2005). At the transcriptional level, expression of CaMETs and CaCMTs showed dynamic changes associated with ripening (fold-change >2), whereas expression of CaDRMs was barely altered (foldchange <2), especially after the MG stage ( Fig. 2C-F), suggesting the CaMETs and CaCMTs may mainly control the maintenance and de novo methylation of cytosine, respectively, during pepper fruit ripening. Moreover, most of the CaMETs and CaCMTs were down-regulated after the MG stage (Fig. 2C, D), suggesting decreased DNA methylation activity. However, similar to tomato, pepper fruit generally undergo genome endoreduplication after the MG stage (Cheniclet et al., 2005;Chevalier et al., 2007;Ogawa et al., 2010Ogawa et al., , 2012. Therefore, passive DNA demethylation is highly likely to occur during pepper fruit ripening. On the other hand, expression of CaDML2, an ortholog of ROS1 (AtDML1) and the highest expressed member of the CaDML family, peaked at the T stage (Fig. 2F), suggesting that an active demethylation process may be occurring during pepper fruit ripening. Thus, it seems that the ripening-associated DNA hypomethylation in pepper may be mainly controlled by CaMETs, CaCMTs, and CaDML2 at the transcriptional level. Similar phenomena have been observed in tomato, in which the DNA hypomethylation associated with ripening is controlled by the expression of both DNA methyltransferase and demethylase genes (Teyssier et al., 2008;Liu et al., 2015;Lang et al., 2017;Yang et al., 2019).
However, in strawberry and sweet orange, changes in ripeningassociated DNA methylation are only attributed to variations in the expression of RdDM-related and DNA demethylase genes, respectively (Cheng et al., 2018;Huang et al., 2019). Therefore, it can be concluded that the transcriptional regulatory mechanism of the variations in DNA methylation that are associated with ripening may be not the same among different species.

Interactions between phytohormones and DNA methylation during ripening
Ethylene is considered to be a positive regulator of climacteric fruit ripening (Oeller et al., 1991;Lanahan et al., 1994;Li et al., 2007;Lee et al., 2012;Sekeli et al., 2014). However, in this study, no endogenous ethylene burst was detected during fruit ripening ( Supplementary Fig. S5), suggesting that our pepper inbred line harbors non-climacteric fruit. Consistent with this, compared with green pericarps of the silenced fruit, CaACS was not up-regulated in premature-ripe pericarps (Fig. 4B). A similar phenomenon was also observed by Aizat et al. (2013). This suggests that an endogenous ethylene burst is not indispensable for the non-climacteric ripening of pepper fruit. In accordance with this, expression of CaPSY1, CaPDS, and CaPG did not clearly respond to short-term treatment with ethephon ( Fig. 5A), suggesting that these genes may not be directly regulated by exogenous ethylene. However, a previous study determined that continuous treatment with exogenous ethylene promoted non-climacteric fruit ripening (Fox et al., 2005), suggesting that exogenous ethylene may have some slow or indirect effects on the ripening of this type of fruit. In our study, expression of CaDML2-like was clearly induced by short-term ethephon treatment (Figs 5A, Fig. 6), suggesting that this gene and even DNA methylation may be involved in the process.
ABA has been widely accepted as an important regulator of non-climacteric fruit ripening (Deruère et al., 1994;Jia et al., 2011;Leng et al., 2014;Wang et al., 2013Wang et al., , 2017. Consistent with this, we found that exogenous ABA promoted carotenoid biosynthesis, cell wall degradation, and DNA hypomethylation through altering the expression of CaPYS1, CaPG, and genes related to DNA methylation (CaDML2-like and CaCMT4like), respectively (Figs 5A, Fig. 6). Endogenous ABA dramatically increased from the MG to R stage (Fig. 5B), suggesting a positive role in the regulation of pepper fruit ripening. ABA levels as well as the expression of CaNCED1 and two CaPP2Cs were up-regulated in premature-ripe pericarps (Supplementary  Table S9, Figs 4B, Fig. 6). CaNCED1 encodes a key enzyme in ABA biosynthesis Tan et al., 1997;Burbidge et al., 1999). CaPP2C encodes a negative component of ABA signaling and is usually up-regulated by ABA (Ma et al., 2009;Sun et al., 2010;Wang et al., 2015). Therefore, it can be deduced that ABA biosynthesis and signaling were activated by DNA hypomethylation in the non-climacteric pepper fruit. Similar phenomena have also been observed in strawberry, in which ABA biosynthesis genes are up-regulated during fruit ripening whilst at the same time undergoing a loss of DNA methylation in their 5´-and 3´-regulatory regions (Liao et al., 2018;Cheng et al., 2018). Hence, it can be concluded that ABA and DNA methylation may interact with each other at the transcriptional level during pepper fruit ripening.
Auxin is often considered as a repressor of non-climacteric fruit ripening that acts fully or partially through antagonizing the effects of ABA (Manning et al., 1994;Davies et al., 1997;Liu et al., 2005;Böttcher et al., 2011;Symons et al., 2012;Ziliotto et al., 2012;Liao et al., 2018). In accordance with this, we found that exogenous IAA inhibited cell wall degradation, ABA biosynthesis, and DNA hypomethylation through affecting the expression of CaPG, CaNCED1, and DNA methyltransferase genes (CaMET1-like1 and CaCMT3-like), respectively (Figs 5A, Fig. 6). Moreover, endogenous IAA levels decreased from the IM to T stage, suggesting that it may be involved in the regulation of ripening of the non-climacteric pepper fruit. In premature-ripe pericarps, the endogenous IAA level as well as most of the differentially expressed AUX/IAA genes were down-regulated (Supplementary Table S9, Fig. 4B). AUX/IAA genes are important elements of auxin signaling and positively respond to auxin treatment (Waseem et al., 2018). Therefore, it can be concluded that DNA hypomethylation caused by the silencing of CaMET1-like1 led to the inhibition of auxin signaling and a decreased level of auxin.

Conclusions
A hypothetical model is proposed (Fig. 6) that indicates a high likelihood of a role for DNA methylation in the regulation of fruit ripening in non-climacteric pepper, which is mainly controlled by CaMET1-like1, CaCMT2like,and CaDML2-like at the transcriptional level. Silencing of CaMET1-like1 by VIGS results in DNA hypomethylation and premature fruit ripening, which includes the accumulation of carotenoids, increased levels of SSC and ABA. and decreased levels of IAA. Under in vitro conditions, ABA promotes DNA hypomethylation in the pericarp at the MG stage at the transcriptional level, whereas IAA inhibits this process. Exogenous gibberellin and cytokinin show indistinct roles in DNA methylation in MG pericarps of non-climacteric pepper fruit.

Supplementary data
Supplementary data are available at JXB online. Table S1. Primers used for McrBC-PCR. Table S2. Primers used for bisulfite sequencing. Table S3. Primers used for real-time PCR. Table S4. Primers used for subcellular localization plasmid construction. Table S5. Primers used for VIGS construction and detection. Table S6. Standard substances used in study. Table S7. IDs and conserved domains of pepper DNA methyltransferase genes. Table S8. IDs and conserved domains of pepper DNA demethylases genes. Table S9. ABA and IAA levels in premature-ripe and green pericarps of CaMET1-like1-silenced fruit. Fig. S1. Upstream regions of transcriptional start sites for McrBC-PCR and bisulfite sequencing. Fig. S2. Subcellular localization of CaDML2-like, CaMET1-like1, and CaCMT3-like in onion epidermal cells. Fig. S3. PCA of the RNA-seq data. Fig. S4. Enrichment of DEGs between premature-ripe pericarps of CaMET1-like1-silenced fruit and red-ripe pericarps of negative control fruits. Fig. S5. Ethylene production of fruit at different developmental stages.
the Institute for Plant Breeding, Genetics & Genomics, University of Georgia, for help with the modification of language. This work was supported by the National Key Research and Development Program of China (2017YFD0101903 to LS and HS), the Beijing Fruit Vegetables Innovation Team of Modern Agricultural Industry Technology System (BAIC01-2018 to HS), and The Construction of Beijing Science and Technology Innovation and Service Capacity in Top Subjects (CEFF-PXM2019_014207_000032 to LS and HS).